Epidemiologic studies are limited in their ability to determine causation; this is particularly important in the case of mixed air pollution where one measured pollutant may also serve as a marker of other confounding exposures (another unmeasured pollutant from the same source, for example).