10/24/94 Please find in the attached zipped files final peer reviews of the Motor Vehicle-Related Air Toxics Study. The five peer reviewers were Tadeusz E. Kliendienst, Andrew Sivak, Thomas H. Stock, Jaroslav J. Vostal, and Clifford P. Weisel. Also included is a summary of the most significant comments received from the peer reviewers. These reviews will be included in the docket for the Motor Vehicle-Related Air Toxics Study (docket no. A-91-19). At this time, we do not plan to revise the study based on these comments. The reviews are being provided to the Regulation Development and Support Division (RDSD) as input for the air toxics rulemaking. Please contact Rich Cook (313-741-7827) if you have any questions or comments. =================================================================== Summary of Peer Reviews of Motor Vehicle-Related Air Toxics Study 1) Tadeusz E. Kliendienst a) Recent research indicates that oxidation of hydrocarbons leads to formation of products (e.g., peroxyacetyl nitrate, PAN) that are substantially more mutagenic than precursors. Also, NOx limited systems tend to be less mutagenic. The significance of these observations with respect to human health impacts is unknown. b) In EPA's urban airshed modeling of benzene in St. Louis, there is a large discrepancy between modeled and measured values of benzene. Possible explanations for this discrepancy should be considered. c) In EPA's urban airshed modeling of 1,3-butadiene in St. Louis, there is a rapid rise in 1,3-butadiene after 20:00 hours LST when NOx levels are highest right after sunset, and 1,3-butadiene reacts relatively rapidly with NO3. This apparent inconsistency should be addressed. d) It is important to note that at the present time, ambient concentrations of toxic compounds cannot be accurately predicted based solely on emission rates, atmospheric dispersion, chemical removal and formation, etc. At the present time, experimental measurements of toxic compounds are the most reliable means of obtaining ambient concentrations and they must certainly be used to validate air quality models for toxic compounds. 2) Andrew Sivak a) EPA should include three recent human studies relating benzene exposure and leukemia in its benzene risk assessment. b) Since there are non-linear dose-response data for benzene, it is unclear how the Agency can state that they support a theoretical linear low dose extrapolation. c) EPA should have used the 1991 draft unit risk for formaldehyde, based on monkey and rat DPX data, in its calculations, rather than the 1987 unit risk. d) EPA should have used newer and more rigorous NTP carcinogenicity data for 1,3-butadiene to calculate unit risk estimates used in the study. e) EPA should have included more discussion of the mutagenicity of polycyclic aromatic hydrocarbons and nitrated polycyclics in the gaseous phase of diesel exhaust. f) The use of a linearized model for the calculation of risk from particle exposure is simply wrong scientifically. g) There appears to be no cause for a public health concern for cancer from exposure to motor vehicle exhaust, and it would seem that the agency could use its resources more productively on matters of higher concern. h) The Agency should consider recent publications by Adam Finkel and others on communicating uncertainties in risk assessment calculations. 3) Thomas H. Stock a) Although use of HAPEM-MS is a laudable first effort in assessing the health impact of exposure to air toxics from motor vehicles, it has many severe limitations which may render it inadequate for its intended use. b) Although HAPEM-MS is a national exposure model, the microenvironmental exposure factors it uses to convert fixed-site concentrations to microenvironmental concentrations are based on data from only one city, Denver. Thus, there are many questions on how much these data can be used to generalize to other cities. c) EPA developed an integrated exposure adjustment factor based on activity data from a California study. It is questionable how representative this is of the nation as a whole. d) Given the limitations of the HAPEM-MS model, it is reasonable to "correct" modelled exposures to agree with adjusted ambient data, until a more realistic model is developed. e) At this point in time, there are insufficient data on any air toxic to explore the relationships among ambient, microenvironmental, and personal exposure. 4) Jaroslav J. Vostal a) Any new assessment of the potential public health risk of diesel particles should include recent data which indicate carcinogenicity of diesel exhaust is not associated with particle-bound organics. b) Tumor responses to diesel particles are not linear at low ambient levels; thus, applicability of the linear dose-response model is restricted. c) Since existing epidemiological data for diesel exhaust exposure does not include documented exposure, the use of epidemiological data in the risk assessment remains questionable. d) EPA's default model of linear multistage extrapolation has no universal validity and needs to be replaced with more realistic approaches once mechanisms are better understood. f) Reanalysis of study results should take into account EPA's new risk assessment guidelines. g) The application of threshold-displaying models may reduce risk estimates for diesel particulate matter, formaldehyde, acetaldehyde or gasoline vapors to levels not statistically different from zero. The same mechanisms may potentially apply also for benzene. This leaves 1,3-butadiene as the only measurable risk of the health effects of air toxics from mobile sources. h) The magnitude of uncertainties and numerical results should be presented in the form of frequency distributions or iterative computer runs rather than by point values. Also, sensitivity analyses should be conducted on the variability and distribution of different ambient concentrations or exposure estimates and the final results should be presented as ranges starting from zero up to the upper bounds. 5) Clifford P. Weisel a) EPA's HAPEM-MS model is severely limited by extrapolating ambient CO measurements to microenvironmental CO measurements nationwide from a single study done in Denver during four months of the winter. These extrapolations need to be validated in other regions of the country in other seasons. b) Time activity pattern data for areas other than just Cincinnati need to be included in the HAPEM-MS model. c) The HAPEM-MS model included five microenvironments, but others, such as public and private parking garages, need to be included. d) It is problematic to extrapolate air monitoring data collected for CO to air toxics because: (1) CO is purely an exhaust emission, while some air toxics such as benzene have contributions from both exhaust and evaporative emissions; (2) atmospheric reactivity of CO and some air toxics is very different. e) Exposures in microenvironments should be considered in calculating cancer risks, since these exposures can contribute a measurable percent of the total daily exposure. f) The inclusion of more accurate estimates of microenvironmental exposure data are expected to produce higher exposure estimates than presented in the study and to reduce the uncertainty of the estimates.