Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 30
Review of the US Navy's Human Health Risk Assessment of the Naval Air Facility at Atsugi, Japan Other Monitoring Although the NEHC draft summary report does not mention monitoring of the incinerator facility with a video camera or the use of optical pyrometers, a supporting document (Radian 1998a) does (referring to an infrared pyrometer initially and in the equipment inventory). The results obtained with those monitors are almost undocumented in any of the reports. The one documented result is the on-off status of the incinerator, which is recorded for each sample in one of the data files provided to the subcommittee and in Table 2-3 of the Radian report (2000a). In addition to indicating the status for each sample, it would be more appropriate to indicate the status hour by hour to correlate with the continuous air-monitoring data. How the monitors (camera, pyrometer, and so on) were used to determine on-off status and any uncertainty involved in that determination should also be described. The subcommittee recommends adding information on how on-off status of the stacks was determined before the optical pyrometer began operating in October 1998 (Radian 1998d, 1999c), including information on the reliability of the method(s). There did not seem to be any analysis of the time-lapse video records. That is particularly surprising in light of the emphasis in the planning stage on the analysis of the tapes to obtain information on plume behavior and fumigation conditions (Radian International, unpublished data, January 27, 2000 4, July 1998 5). The comparison of the Atsugi mean contaminant levels with the US mean values in Table 2.5 (NEHC 2000; pp. 25-26) is not valid. The values used for comparison in that table are presented as US mean values but are not. They represent data collected in a small survey in California or very old exposure estimates reported by Shah and Singh (1988) that are not representative of average US exposures to the substances in question but are averages of all reported indoor-exposure measurements. APPENDIX F Health Risk Assessment Human-Health Risk-Assessment Results The results section of NEHC's draft summary report (2000, pp. 35-40) is an abbreviated version of the Pioneer (2000) document. The section addresses potentially exposed populations at NAF Atsugi and presents health-risk estimates for children and adults exposed for 3, 6, and 30 yr. Without a complete description of the population at risk, however, it is difficult to evaluate the relevance of those exposure scenarios. (For example, are subgroups of the population at greater risk, such as military personnel and their families who have repeated, but nonconsecutive and therefore not limited, tours of duty at NAF Atsugi?) Various direct exposure pathways are introduced and discussed. Indirect exposures are not considered, because most food is assumed to be supplied from the United States and the drinking water is assumed not to be contaminated by incinerator fallout. Those assumptions should be better documented with supporting data. In addition, the fact that drinking water is not affected by the incinerator facility does not obviate the assessment of drinking-water contaminants to determine the overall health effects of residing at NAF Atsugi. 4 Radian (Radian International). 2000. Response to comments draft sampling and QA/QC plan to assess health risks related to air quality at NAF Atsugi, Japan. Provided to subcommittee on CD-ROM, file qapp_comments.doc. Dated January 27, 2000. 5 Radian (Radian International). Review comments first quarterly report 21 April - 31 July 1998. Provided to subcommittee on CD-ROM, file nech_comments 1st qtr.doc. Dated January 28, 2000.
OCR for page 31
Review of the US Navy's Human Health Risk Assessment of the Naval Air Facility at Atsugi, Japan As is appropriate, the risks of cancer and noncancer effects are discussed in the NEHC draft summary report. It states that the cancer risk for children is “slightly higher” than the EPA benchmark (10−4), as is the risk of noncancer effects. Table 3-2 of the NEHC draft summary report indicates that 60% of the exposure scenarios for children have cancer risk estimates that exceed 10−4, compared with 15% of adult exposure scenarios. Higher noncancer-hazard indexes are also observed for children than for adults for every exposure scenario except recreational golfers. The source of the apparently larger risks for children should be clearly identified by NEHC. For example, are the differences from adult risks due to higher soil-ingestion rates, breathing rates per unit body weight, and so on? With respect to the risk of noncancer effects, the document gives the impression that risk is related largely to respiratory effects and that those effects are reversible, but both impressions might be overstated. Several of the chemicals of concern are reported to be reproductive and developmental toxicants; those types of toxicity should be given more consideration. Reproductive and developmental effects of dioxin, for example, have recently been reviewed (Yonemoto 2000). In addition to listing the hazard index in Table 3-2 (p. 37; NEHC 2000), it would be useful to indicate whether any hazard quotients exceed 1. The meaning of notes to Table 3-3 (p. 40; NEHC 2000) is not clear. Health Evaluation On p. 41 of the NEHC draft summary report, the stated purpose of the health-evaluation section is “to interpret and provide a context for presenting the results of the risk assessment, to discuss health concerns associated with the risk and to further characterize the estimated health risks based upon the sampling and risk analysis performed at NAF Atsugi.” Overall, much of this section is repetitive of earlier sections of the draft summary report and not central to its stated purpose. Basic questions for persons residing at NAF Atsugi are how the incinerator is affecting their health and how certain NEHC is about the effects; for example, “How many studies have been completed on Jinkanpo and who did the studies?” (see “Frequently Asked Questions ” in Appendix B, NEHC 2000). Those questions are not adequately answered in the health-evaluation section. The subcommittee recommends that some attention be given to the risks for susceptible subpopulations (such as children with asthma, preterm infants, and pregnant women). The text pertaining to children's risks (NEHC 2000; p. 55) is vague and superficial and does not consider potentially susceptible populations. Given NEHC's emphasis on using EPA methods, it is surprising that it does not follow EPA's increasing focus on childhood risks. For example, NEHC does not fully address exposure of infants to dioxin through breast milk. A recent paper by Patandin et al. (1999) illustrates that high levels of dioxins can be transferred by this pathway, and failure to consider it is inconsistent with published EPA methods (EPA 1998a). NEHC also does not address the potential initiation, exacerbation, or persistence of asthma due to chemical or particle exposures (see reviews by Jones 2000; D'Amato 1999; Goldsmith and Kobzik 1999; Linn and Gong 1999). The calculated cancer risk estimate is an upper bound on lifetime probability of developing cancer under defined exposure conditions. NEHC uses the RME to estimate an upper bound on the estimates. If a number of upper-bound estimates of exposures are used to estimate risk, then on the basis of simple joint-probability calculations for independent events, the estimated risk will most likely be much higher than the actual risk. The same logic applies to the average-exposure scenario; the probability outcome of multiple mean estimates is unlikely to be an average result and more likely (in these types of risk assessments) to be an upper percentile, depending on the number of separate variables and on details of the distributions. The nature of the cancer risk and exposure scenarios should be taken into account in the risk assessment and its interpretation. An alternative approach would be to use a more sophisticated distributional analysis that could
OCR for page 32
Review of the US Navy's Human Health Risk Assessment of the Naval Air Facility at Atsugi, Japan incorporate both individual variability and uncertainty. Cullen and Frey (1999) present more information on that type of analysis. EPA's Integrated Exposure Uptake Biokinetic (IEUBK) model for lead is calibrated on the basis of arithmetic or geometric averages as input values. The NEHC draft summary report (p. 69) suggests that RMEs were used as input into the IEUBK model for this risk assessment, implying the use of 95th percentile upper confidence bounds (UCL95) on the means for air and soil concentrations, and the upper end estimates for other parameters as selected for the RME in the risk assessment; that would not be appropriate. The Pioneer (2000) draft report, however, suggests that estimates of UCL95 on the means for air and soil concentrations were used as input into the model, and that the IEUBK default values were used for other parameters; these also are not ideal inputs for the model. A preferred approach is to use median estimates for exposure concentrations as point-estimate inputs into the IEUBK. Evaluation of the variability among individuals would require a convolution of the variability distributions for the exposure-point concentrations with the lognormal variability distribution included with the IEUBK to estimate variation among individuals exposed to fixed input concentrations. The subcommittee also notes that the value of 3.9 µg/m3 used in the risk assessment (Pioneer 2000) as the UCL95 on the sitewide mean air lead concentration is incorrect by a factor of about 10: the UCL95 on the mean is close to 0.4 µg/m3, although the estimate depends somewhat on the assumptions made about the distribution. On p. 69, NEHC's discussion of lead measurements states that “this value is well below the Centers for Disease Control and Prevention (CDC) benchmark of greater-than-5 percent probability.” The CDC has no such benchmark. As NEHC correctly indicated in the previous paragraph on p. 69, the 5% probability is a guidance level from EPA. Page 54 of the NEHC draft summary report states “typically, RfCs are one-thousandth of a NOAEL; therefore, a hazard index of 10 would be acceptable in these cases because there would still be a safety margin of exposure of 100.” That statement confuses the basis of uncertainty and modifying factors and of their relationship to the hazard index. Uncertainty and modifying factors are used to extrapolate to safe exposure levels for humans, accounting for uncertainties resulting from differences between studied exposures and possible human exposures. The average human might be ten times more susceptible than the average member of the most susceptible animal species studied; a highly susceptible human might be ten times as sensitive as the average; and human exposure can be ten times higher than the longest exposure observed in the laboratory. Similarly, the term “safety factors” as used by NEHC (p. 54) is not appropriate. It is also unclear whether NEHC is attempting to define a universal value for the hazard index that would correspond to a point where health effects might be expected or to define an acceptably low value of the hazard index to dismiss all concerns about health effects. Clarity is critical to the question of what the overall goal of the risk- assessment project is. Is it attempting to show that there is no problem, or is it attempting to see whether there is a problem? In the first paragraph on p. 75, citations should be added for all the concentrations and locations presented. In the second and third paragraphs on p. 75, the information on EPA and the RfC for acetaldehyde is repetitive and contradictory. The RfC given in the second paragraph is incorrect. The correct value is given in the third paragraph. The discussion of acrolein on p. 76 is simplistic and confusing. The effect of acrolein on nasal epithelium discussed in the second sentence of the third paragraph on p. 76 is most likely a portal-of-entry effect, not a systemic effect, as indicated in that sentence. On pp. 69-77, the brief descriptions of the health effects of various chemicals are not clear and do not add to the document. It would be preferable to include an evaluation that incorporates known and suspected adverse human health effects.
OCR for page 33
Review of the US Navy's Human Health Risk Assessment of the Naval Air Facility at Atsugi, Japan Pioneer Risk-Assessment Document On p. 7 of the Pioneer (2000) draft report, it is stated that hole 9 of the golf course “frequently receives emissions from the incinerator stacks.” However, the wind rose indicates that the wind is from west, west-southwest, or west-northwest about 2.7% of the time (average, about 10 d/yr). The statement on p. 8 of the Pioneer (2000) draft report that “these assessments do not address risks from other sources of exposure (e.g., dietary exposures) or risks from other constituents that are not associated with the site under evaluation” is also not consistent with the first objective of estimating the potential human health risks at NAF Atsugi. On p. 12 of the Pioneer (2000) draft report, it is stated that a 0- to 3-in. deep (0-7.6 cm deep) soil sample was used “because it is representative of the portion of the soil column that most people routinely contact.” However, people do not routinely come into contact with soil below the surface layer down to a depth of 3 in. (7.6 cm). Soil samples up to 3 in. (7.6 cm) deep might provide the closest available surrogate for the soils that people actually come into contact with. In some circumstances (such as longer exposures), if there is sufficient mixing of surface soil through this depth range for the concentrations in the entire depth range to be of relevance, those soil samples might be appropriate. Discussion of the potential mixing rate of surface soils, its effect on the soil-contact scenario, and the collection of surface-only samples (the top millimeter or so) should be considered in the planning of future studies. As discussed on p. 14 of the Pioneer (2000) draft report, duplicate field results are treated as independent observations when calculating summary statistics and exposure-point concentration estimates. Such treatment doubles the weight placed on the concentration at a single place and time, and is not appropriate. Field or laboratory duplicates should be averaged before summary statistics are calculated. Half the quantitation limit (QL) was substituted for the concentration of all nondetected values, provided that a chemical was detected somewhere, in the calculation of exposure-point concentrations (Pioneer 2000; p. 14). Hornung and Reed (1990) compared three methods for calculating the mean for censored, lognormally distributed concentration data, for 20 different combinations of the percentage of data below the limit of quantitation and geometric standard deviation (GSD). The most accurate for all conditions, but also the most difficult to use, was the maximal likelihood method of Hald (1952). That method cannot be used when over half the data are below the limit of quantitation. The two simpler methods considered were the substitution of QL/2 or QL/√2 for measurements below the limit of detection. The QL/√2 method was better than the QL/2 method for low to moderate variability (GSD < 3), but not as good for GSD = 3. NEHC should include a justification for the use of the QL/2 method to strengthen the discussion of the risk assessment with mean concentrations. Page 14 of the Pioneer (2000) draft report mentions a procedure called “Compound Rules of Decision”. The procedure is neither described nor referenced, and it is not stated whether the circumstances under which it is supposed to be invoked ever occurred in the risk assessment. Some description of the procedure, the specific circumstances under which it was invoked in this risk assessment, and a citation should be included. Section 2.3 (Pioneer 2000; p. 15) describes the initial screening of chemicals of concern (COCs). Such a screening, if carried out as stated, would prevent the risk assessment from addressing its first objective, because the overall risks of the site would include those due to background concentrations. It appears that the screening was carried out for the soil measurement but not for the air measurements. The implications of that should be discussed. On p. 16 (Pioneer 2000), the descriptions of the calculation of the UCL95 estimates of mean concentration are not clear—for any distribution, an estimate of the UCL95 of the mean is required. How a distribution is tested for normality or lognormality is not specified, nor are the criteria applied to the results of any such test. If the distribution is neither normal nor lognormal, further analysis might be desirable before an approach based on normality is accepted. The description is also inadequate in that the estimates adopted
OCR for page 34
Review of the US Navy's Human Health Risk Assessment of the Naval Air Facility at Atsugi, Japan for the UCL95 of the mean are not given. Many estimation procedures are available (such as, analytic estimates based on the t-distribution for normals and on Land's procedure for lognormals, minimum variance unbiased estimates, likelihood-based estimates, and bootstrap and jackknife estimates applied to any of these or others), and the procedure used should be stated. Page 22 (Pioneer 2000) describes the term MF (defined by Pioneer as exposure-pathway- and constituent-specific modifying factors, such as percutaneous absorption rate) in the first equation is described as having “variable units”. With the definitions given for the other variables in the equations, MF is dimensionless. Page 22 (Pioneer 2000), has the following explanation of how the exposure parameters were chosen for the RME case: Each variable in this equation has a range of possible values associated with it. The intake variable values for a given pathway are selected so that the combination of all intake variables results in a realistic upper bound estimate (or RME) of the possible exposure by that pathway. The same values, however, were used for the average case. In the risk assessment, the average case appears to use RME estimates for all exposure parameters except the exposure-point concentration, where the difference between average and RME cases is the difference between an estimate of mean concentration and a UCL95 on the mean concentration (see, for example, Tables 3-2 through 3-6, particularly their footnote c). That is not the usual meaning of average for such exposure scenarios and is misleading. The typical approach for estimating an average or central-tendency case is to obtain an average for the whole population that is exposed by using exposure parameters that represent central-tendency values (such as means or medians). The ranges or confidence limits around the central-tendency values should also be presented. In Table 3-2, footnote d confuses the “fraction from contaminated source” with “outdoor and indoor exposure to soils”. Although those concepts might overlap in some circumstances, they are distinct and do not overlap in this case. The formula presented appears to have been adapted in such a way that the “fraction from a contaminated source” represents the “fraction of time indoors”. The explanation in the table should explain that better. In Table 3-3 (p. 23), footnote e does not explain how 150 mg/d is “the midpoint” between 50 and 200 mg/day. Pages 33 and 68 (Pioneer 2000) mention an inhalation RfD, the second time in the context of EPA's IRIS database. That RfD was probably derived from an RfC. The term “inhalation RfD,” however, is not used by EPA and is confusing. Page 84 (Pioneer 2000), in the context of a comparison between the golf-course site and the GEMB site, states that “the only difference between the airborne concentrations, and consequently risk, at the GEMB and the airborne concentrations at the Golf Course should be emissions associated with the SIC.” That would be correct only if the “Background + Other Point and Non-Point Sources (emissions)” affect the two sites equally. That hypothesis was not established or tested at any point in the project. On pp. 84-85 (Pioneer 2000), the methods adopted for the comparison between the golf-course and GEMB sites are not adequately explained. For example, there is no information in the documentation as to which particular days were used for the comparison. Even if the days were correctly selected, the results presented in Table 5-10 cannot be interpreted without further information on the method because some approaches to producing such values are statistically invalid. Consider, for example, the values in the columns labeled “RME” in Table 5-10. “RME” is not defined in this context, so the subcommittee had to infer that it has some connection with an upper 95th percentile estimate of an average because that was the apparent previous meaning of this term in the document (as given on pp. 15-16, Pioneer 2000); but there is no indication of how this 95th percentile was calculated. The committee can conceive of several ways of generating the values in Table 5-10. For example,
OCR for page 35
Review of the US Navy's Human Health Risk Assessment of the Naval Air Facility at Atsugi, Japan Take the concentrations measured at each site for each chemical on the selected days, and find the average and an upper 95th percentile estimate on that average for the concentration of each chemical at each site. Calculate an “average” and “RME” risk estimate for each site on the basis of the two sets of concentration estimates. The entries in Table 5-10 could then be the differences between sites. Although the “ average” estimate so generated for a site is meaningful, the “RME” site differences so obtained have no statistically valid meaning. This method appears to be the closest approximation to what was meant by “RME” in the rest of the document; but the differences between such “RME” values cannot be interpreted. Take the concentrations measured at each site for each chemical on the selected days, calculate day-by-day concentration differences for each chemical, and compute the average of these differences in daily concentrations for each chemical over all days selected and upper 95th percentiles on such average differences. Calculate the risk estimate differences for Table 5-10 on the basis of the two measures (“average” and “RME”) of concentration differences. The “average” so obtained will be the same as for approach 1, but the “RME” value will be different and will have no statistically valid meaning. For each selected day, calculate at each site a risk-weighted sum of concentrations of all the chemicals in question, selecting the risk weighting so that summing over all days would give a risk estimate (roughly speaking, a risk estimate for that day for that site). Take the difference between the values for each site to obtain a series of daily risk-weighted differences. Obtain the sum and the upper 95th percentile estimate on the sum of the risk-weighted differences as “average” and “RME” estimates. The “average” value so obtained will be the same as approaches 1 and 2, but the “RME” will be a statistically valid estimate that can be interpreted. The statistical uncertainty associated with the “average” column in Table 5-10 is not presented—the differences might not be statistically distinguishable from zero. The third of the approaches just summarized provides a series of daily values that would allow calculation of statistics on differences between sites, including the statistical significance of such differences. In contrast, the first and second approaches (and many other possible ones) cannot provide such information. In any case, the values in Table 5-10 cannot be used to draw unequivocal conclusions about the contribution of the incinerator without an evaluation of the hypothesis that there is no difference in the absence of the incinerator. Moreover, such values as those in the table would allow an estimate only of the contribution of the incinerator to the differences between the GEMB and golf-course sites, not of the average contribution to actual populations or individuals. It is pointed out on pp. 85-86 (Pioneer 2000) that the majority of the hazard-index estimates is contributed by acetaldehyde, acetonitrile, acrolein, and PM10, but it is not pointed out that of those major contributors, only PM10 could be associated with incinerator emissions in the analyses presented. The “upwind” hazard index or risk estimate is higher than the “downwind” for several chemicals in Tables 5-11 and 5-12; that situation would not be possible (except for the inherent uncertainties) if the hypothesis that there are no differences in the absence of the incinerator is correct. Although such effects could be due to the uncertainties involved, the uncertainties are not discussed. The recommendations section of the Pioneer risk-assessment document (Pioneer 2000) contains many recommendations that are not based on the findings and conclusions presented in the draft report. The primary recommendation (recommendation 1, p. 92) mentions specific periods (32 and 98 months) that are not mentioned in the draft report. Recommendation 2 (p. 93) is common sense, but no part of the draft report has any evaluation of the effectiveness of the recommended actions. Recommendations 3, 5, 6, 7, and 10 also bear no relationship to any analysis in or conclusion in the draft report.
Representative terms from entire chapter: