US EPA

If ambient PM is toxic, a statistical relationship might be expected between personal exposure to ambient PM and health effects due to ambient PM. However, some exposure analysts seem to believe that there cannot be a meaningful relationship between ambient concentrations of PM and health effects due to ambient PM exposure unless there is also a strong relationship between ambient concentrations of PM and total personal exposure to all PM from all sources. They consider it paradoxical that statistically significant relationships are found between ambient PM concentrations and health effects even though the correlations between ambient PM concentrations and personal exposures to all PM are low and frequently near zero. Some authors have expressed the opinion that this paradox raises questions regarding the scientific basis or effectiveness of EPA's PM standards. In an effort to resolve this paradox, authors have unsuccessfully sought explanation by investigating longitudinal studies of individual subjects and by considering the possibility that elderly subjects or subjects with COPD or heart disease might have different exposure patterns. This paper attributes the paradox to two errors in interpretation. One error is use of an incorrect aspect of personal PM exposure. Personal exposure to ambient PM (both while indoors and outdoors), not toal personal exposure to both ambient PM and nonambient PM, should be used in the correlation. The second error, as applied to community time series epidemiology, is an inappropriate selection of data for forming the correlation of total personal PM exposures with ambient PM concentrations. Daily ambient PM concentrations should be correlated with the daily community average of the measured total personal PM exposure of a cohort of subjects, not with the pooled individual total personal exposure values of a group of arbitrarily selected subjects on arbitrarily selected days. Both theory and experiment indicate that either approach will yield much higher correlations and that a combination of the two approaches will yield very high correlation coefficients, above 0.98. This resolves the paradox and indicates that community time-series epidemiology yields information on the health effects of ambient PM, but that other approaches must be used to investigate the health effects of nonambient PM.

Folinsbee, L. Air Pollution And Exercise-Induced Bronchoconstriction. Marcel Dekker Incorporated, New York, NY, volume 130, (1999).