Tobacco Documents Online

UNITED STATES ENVIRONMENTAL PROTECTION AGENCY OFFICE OF RESEARCH AND CLEVELOlPMENT ENVIRONMENTAL CRITERIA AND ASSESSMENT ORF(CE GtNCINNATt. OH10 452156 March 21, 1992 SUBJECT: Review of the Report on Respiratory Effects from ETS FROM: Richard C. Hertzberg~ MEDS TO: Pat Murphy, Review Coordinator MEDS Revi.w Conditions This review only considers section 8 and Appendix B. The severely restricted time allowed for the review precluded the consideration of other sections. The principal methodological issues depend strongly on the validity of the epidemiologic studies, which I am not qualified to judge. Consequently, this review concentrates on procedure: whether major assumptions are listed, Judgments are presented on tha validity of each assumption, calculations are correct given the presumed validity of the assumptions, and conclusions are characterized within the context of the assumptions and variability identified earlier in the report. chapter 8 The understanding of this chapter would be aided by a summary table, for each stage or subsection, of assumptions, their validity, and their numerical impact on the concluding risk sstimates. Also helpful would be a table listing the ma jor sources of uncertainty in ths final risk estimates and, for each source, whether it was possible to include a numerical representation of that uncertainty into the final range for the risk estimate. Section 8.3 The key information on which calculations are based is cotinine concentrations in body fluids. Yet this measure is sharply criticized (top of p. 8-6) and the warning is given that dependent results "may thus be subject to considerable error and should be interpreted with caution." This uncertainty is not addressed in the ensuing calculations, but should be. In several ~ instances, only mean cotinine values are given, and in others an ~ undefined range is presented. My interpretation of this section is ~ that the various studies cited, combined with the adjustment model ~ (eq. 8.1), are only used to indicate the potent3a.I tor'~ underestimation of the true relative risk but that no numerical CP correction tor that underestimate will be performed. Tf this ~ impression is correct, it should be stated more clearly. .._. ~ Pnnt®d on Recy:led Paper

The NRC assumptions (bottom of p. 8-5) actually carry an additional assumption that should be stated. The first two listed are: [cot) - a, + b1* [ETS„bi,.] excess riskre,p - a2 + b2* [ETS,w,,,bd] Clearly, for cotininn to be an indicator of ambient ETS exposure, there needs to be a relationship between ambient ETS and absorbed ETS, or between internal cotinine and absorbed ETS. The later use of this information (eq. 8.1) implies this "absorption" model is linear, but it should be so stated as an additional assumption. The discussion following eq. 8.1 is confusing. The left-hand side is the observed RR for individuals reported as "exposed." The right-hand side uses dN for individuals reported as "unexposedi' that includes those truly unexposed in addition to those misclassified. The conclusions (p. 8-7, para. i),, however, give bdN as the increased risk in misclassfffed persons and then give the RR for the "unexposed" person (both misclassified and true unQxposed). I believe the bdN term actually applies to the entire group designated as "unexposed" since that is the group from which the Z factor is derived. It then represents the average excess risk for the group, not just for the misclassified persons. The adjusted RR for the exposed-classitied child, given as 4, then seems to be correct, but no inference seems possible for those in the "unQxposed" group that are misclassified. It follows that the misclassified "unexposed" persons have RR > 2. This discussion might be easier to follow if you identify the excess risk terms, e.g., change para. 1, line 6 to be: Solving for the excess risk in the "unexposed" category, bdN - 1. Note that the following phrase should be "twice the observed relative risk" instead of observed risk. Section 8.4 in para. 2, line 6-7 ("This may explain, for...."), the phrase "stronger relation" is misleading. The previous sentence suggests that the number of parent-smoked cigarettes is a poor and highly uncertain indicator of total ETS exposure. The relation you identify is then not stronger in the sense of smaller variation (a better model) but shows a st.eper slope (e.g., x cigarettes corresponds to 3y disease cases instead of y, due to the extra BTS unaccounted for by the parents' cigarettes). The uncertainty could still be very high. Somewhere you should carefully define what you are estimating, I believe it is excess risk according to the definitions of Greenland and Robins1 (1988), i.e., only those cases that would not have occurred had exposure not occurred. The cases excluded, that ' Greenland, S. and J. M. Robins 1988. Conceptual problems in the definition and interpretation of attributable fractions. Am J Epidemiology 128(6):1185-1197.

you should note, are those that would have occurred without the Qxposure, but whose severity is increased or time of onset is advanced because of the exposure (Greanland and Robins' 'ftype 1" case).