On the interpretation of epidemiological studies of ambient air pollution