Review of the Air Quality Criteria Document for Particulate Matter

6. Susceptible sub-populationsIt is surprising that the most important study to date on identifying susceptibility ofpopulations subgroups based on pre-existing medical disorders is discussed so little (Goldberg2000), being presented last in a discussion of previous studies that, because of design, are limitedin the information that they provide in this regard. This study confirms many of the findings ofstudies that attempt to address the issue by stratifying on cause of death. However, it isinteresting that no increased risk was identified for the subgroup of subjects with chronicobstructive lung disease, a group considered to be at high risk based on cause-of-deathstratifications. Parenthetically, I believe the description of the Goldberg study gets it wrong.PM pollutant measures were associated with mortality, not with acute respiratory disease, etc.(p.6-74, line 24-30) as stated. The latter were the susceptibility subgroups (that is, for assessinginteraction effects, essentially).There is legitimate concern that the stratification of by cause of death is fraught withproblems misclassifying7. Miscellaneous “large” issuesGaseous pollutant effects: The summaries provide relatively balanced syntheses ofrecent gaseous findings (p.6-75, line 14-23 & p. 6-76, line 24-). This balance is sometimeslacking in the descriptions of specific studies in which a “particle-centric” perspective ismaintained (see point 2 above). For example, the conclusion that fine PM effects oncardiovascular hospitalizations are most important (6-235, line 1) ignores the important findingsby Burnett and Moolgavkar on the role of gases in affecting estimates of PM effects.Threshold concentrations: The discussion of thresholds is unconvincing. The argumentattributed to Schwartz that a threshold is mathematically impossible in the face of populationdifferences in sensitivity (p.6-246, line 3-5) holds only if the most sensitive members of apopulation are sensitive to very low concentrations, which may not be the case. Further, the CDis not consistent in its support of a no-threshold concentration-response relationship. Theargument that heterogeneity in studies of PM composition is due to variable concentrations ofPM components (with studies showing no effects having concentrations too low to show effects,6-78, line 1) is not consistent with the absence of thresholds. The same point can be made ifheterogeneity of effects in NMMAPS is argued to be due to variability in PM concentrationsacross city (see point 4 above).Measurement error: The description of the Zeger (2000) work on measurement error(6.249-252) is just about comprehensible. “Dumbing” this section down, if possible, would beallow it to have the impact that it deserves.Smaller issues for chapter 6 (by page number):Introduction6-2 I don’t believe Rothman would assign more inferential strength to case-control studiesthan cohort studies (line 14).6-3 The prospective cohort studies in this setting do not make use of “individual exposure”(line 2). The subjects in a cohort study do not need to be recruited independent ofexposure (line 4), and in fact were not (e.g., 6-Cities Study).6-4 Line 19-25. The discussion of causal pathways, although correct, is not relevant in thiscontext. Because SO 2 contributes to sulfate formation does not imply that SO 2 effectscannot be separated from sulfate effects, if correlations are not too strong. But, this is anissue of collinearity.6-5 One gets the impression here that meteorology is acting solely as an effect modifier (line1), when in fact the more important issue is its role as a potential confounder (see“Confounding” section above).A - 14