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 1
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar THE NATIONAL ACADEMIES Advisers to the Nation on Science, Engineering, and Medicine Nuclear and Radiation Studies Board 500 Fifth Street, NW 6th Floor Washington, DC 20001 Phone: 202 334 3066 Fax: 202 334 3077 www.national-academies.org/nrsb March 7, 2006 Lt Col Paul Legendre, Chief Bioenvironmental Engineering Office of the Command Surgeon Headquarters Air Force Space Command 150 Vandenberg Street, Suite 1105 Peterson AFB Colorado Springs, CO 80914-4550 Dear Lt Col Legendre: This letter report follows the release of the 2005 report entitled An Assessment of Potential Health Effects from Exposure to PAVE PAWS Low-Level Phased-Array Radiofrequency Energy, prepared by the National Research Council (NRC) Committee to Assess Potential Health Effects from Exposures to PAVE PAWS Low-Level Phased-Array Radiofrequency Energy. That report describes possible health effects of the PAVE PAWS radar located at the Massachusetts Military Reservation in Cape Cod, Massachusetts. This is the fourth and final report to be issued by this committee. Two previous letter reports (NRC 2002a, 2002b) provided advice on the Phase IV waveform measurement effort and evaluate the status of information available to the committee. A third report provided an assessment of potential health effects from exposure to the PAVE PAWS radar (NRC 2005). TASK FOR THE COMMITTEE This committee’s original charge was to provide an update of an earlier report (NRC 1979) and provide a discussion of: the applicability of, and the level of uncertainty associated with, using data derived from cell, animal, and epidemiological studies employing continuous-wave exposure for evaluation of potential adverse health effects following phased-array exposures; the extent of the exposure of the public to electromagnetic energy from the PAVE PAWS system; potential biological and health effects of the PAVE PAWS radar system; and recommendations for appropriate follow-on study-design issues including the strengths and limitations of the approaches suggested and the potential value of the proposed work.
OCR for page 2
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar The committee addressed this charge in its 2005 report (NRC 2005). However, the committee was unable to consider an ongoing epidemiological study being conducted by the International Epidemiology Institute (IEI), which was not available at the time. The committee and the Air Force decided to proceed with the 2005 report with the understanding that the committee would evaluate the results of the epidemiological study in a subsequent letter report, anticipating that the additional information might address several of the questions raised in the 2005 NRC report or might modify conclusions and recommendations provided in that report. This letter report provides the committee’s review of IEI’s epidemiology study. The purpose of this letter report is to: Independently review the International Epidemiology Institute’s (IEI) draft final report entitled A Public Health Evaluation of Radiofrequency Energy from the PAVE PAWS Radar, Cape Cod Air Station, Massachusetts: Descriptive Studies of Disease Occurrence and PAVE PAWS Radar (IEI 2005), and evaluate it in terms of the study design, conclusions, and whether that study made the best use of available data, (the IEI report is an epidemiological evaluation of possible health effects of the pulsed, phased-array PAVE PAWS radar); Review the conclusions reached in the committee’s 2005 NRC report in light of the draft final IEI report; and Review the NRC committee’s recommendations for follow-on epidemiological studies, including the strengths and limitations of the approaches suggested and the potential value of the proposed work in light of the IEI report, and modify the committee’s recommendations if appropriate. 1. INDEPENDENT REVIEW OF THE INTERNATIONAL EPIDEMIOLOGY INSTITUTE DRAFT FINAL REPORT The IEI draft final report contains considerable information not available when the 2005 NRC report was written. That information includes a cancer incidence study of a phased-array radar not reviewed in the committee’s 2005 report (IEI Report, Chapter 1: B-4 Cancer Incidence Study Around the Fylingdales Phased Array Radar Installation); secular trend analyses of county cancer mortality rates (IEI Report, Chapter 2: Secular Trend Analyses of County Cancer Mortality rates); dose-response analyses of mortality and hospitalization rates (IEI Report, Chapter 3: Dose-Response Analyses of Mortality and Hospitalization Rates); cancer incidence rates (IEI Report, Chapter 4: Dose-Response Analyses of Cancer Incidence Rates); and birth weight (IEI Report Chapter 5: Dose-Response Analyses of Birth Weight). The committee reviewed the chapters in the IEI draft report in the order presented. A number of what the committee believes to be factual errors in the IEI report are noted in Appendix A. IEI Report, Chapter 1: B-4 Cancer Incidence Study Around the Fylingdales Phased Array Radar Installation The committee notes that IEI completed a literature review of the potential health effects from radiofrequency exposure and included one additional study not reviewed by this committee in its 2005 report, a descriptive study of exposure to a phased-array radar installationthe U.S. Air Force Fylingdales radar installation in the United Kingdom (NHS 2003). The findings of this descriptive study, which included cases of cancer of the breast, lung, and colorectum, hematologic malignancies, and total cancer incidence between 1991 and 2000, did not provide evidence of harmful effects to the public from low-level radiofrequency radiation exposure. Rates for total cancer in the wards close to the radar were significantly lower than those
OCR for page 3
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar elsewhere in Yorkshire County, and data for specific cancer sites also showed this relationship. We note that the relevance of this study is limited by the lack of exposure data. IEI Report, Chapter 2: Secular Trend Analyses of County Cancer Mortality Rates Selection of Comparison Counties An important issue to consider regarding the secular trend analyses conducted by IEI is the choice of comparison counties, that is, Barnstable County (Cape Cod) versus Berkshire, Hampshire, and Worcester Counties. The strategy for these analyses is that if PAVE PAWS has increased the rate of any cancer, it might be possible to detect this effect without exposure information by comparing cancer mortality rates for counties close to PAVE PAWS to counties farther away. This assumes that all these counties are similar with respect to other predictors of disease rates (e.g., demographic and socioeconomic indicators). Note that without the use of direct exposure information, this set of analyses compares changes in disease rates on Cape Cod with those observed elsewhere in Massachusetts. As such, it is not possible to rule out, as an explanation for observed differences, the influence of other environmental, behavioral, demographic, or socioeconomic changes coinciding with, but not related to, start-up of the PAVE PAWS radar. In conducting this comparison it is important that the counties compared be as similar as possible in terms of demographic, socioeconomic, and other characteristics, so that differences in these factors among the counties do not themselves result in differences in disease occurrence. IEI specified and used an algorithm to select comparison counties from all counties in Connecticut, Massachusetts, and Rhode Island and selected the Massachusetts counties of Berkshire, Hampshire, and Worcester Counties as most similar to Barnstable. However, as noted in an external review included in the appendix to the IEI report, the selection of Worcester County was problematic because it includes “an old industrial town with many relevant characteristics that likely are different from those of Barnstable County” (p. 137). The committee concurs with the external reviewer’s comment and did not see an adequate response in the IEI report regarding this concern. Urban, industrial areas typically have higher cancer rates than other areas, which may distort the intended comparison. To evaluate that possible confounder, the committee reviewed the population size and population density information for the counties from the 2000 U.S. Census. According to the U.S. Census, Worcester City has a population density of approximately 4,500 persons per square mile, some county subdivisions with population densities over 1,300 persons per square mile, and still other subdivisions with a population density of less than 300 persons per square mile. The most densely populated county subdivision contains approximately 17 percent of the total comparison population, giving it substantial influence in the overall results. Other subdivisions in Worcester County also have moderately high population densities (over 1,300 people per square mile). In Barnstable County, on the other hand, the most densely populated subdivision has a population density of 1,023 persons per square mile, while the least densely populated subdivision has a population density of about 100 persons per square mile. Even though the total county population densities for Barnstable and the comparison counties are equivalent, the within-county variation of population density is quite different for Worcester County, which includes a large urban region not present anywhere in Barnstable County. The 1990 census data show that approximately 58 percent of the population in Worcester County lives in urban areas as compared with 38 percent, 38 percent, and 33 percent in Barnstable, Berkshire, and Hampshire counties, respectively. It is possible that sociodemographic and health factors associated with the urban characteristics of Worcester County may contribute to higher disease rates. This suggests the possibility of substantial bias due to selecting Worcester County as a comparison. The two other comparison counties, Berkshire and Hampshire, are more similar to Barnstable County than is Worcester
OCR for page 4
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar TABLE 1 Year 1990 and Year 2000 Populations in Barnstable and Three Comparison Counties 1990 Population Percent Percent urban Percent rural Density (#/mi2) 2000 Population Percent Percent change 1990-2000 Barnstable 186,605 --- 38 36 471 222,230 --- +19 Comparison Counties Berkshire 139,352 14 38 40 150 134,953 13 −3 Hampshire 146,568 15 33 24 277 152,251 15 +4 Worcester 709,705 71 58 29 469 750,963 72 +6 TOTAL 995,625 100 --- --- --- 1,038,167 100 --- Worcester City 169,759 17 --- --- 4,467 172,648 17 +2 County with respect to the proportion of urban residents and distribution of population densities among their subdivisions (see Table 1). To assess the influence of using Worcester as one of the comparison counties, the committee attempted to calculate mortality rate ratios similar to those provided in Chapter 2 of the IEI report, excluding Worcester County. Here, an inconsistency was found that precluded direct comparison. The only tables in the IEI report that provided data for Berkshire and Hampshire counties without Worcester County used a different age standardization1 approach and different time periods of observation. In terms of the standardization, the mortality rates used in Chapter 2 were summarized using indirect standardization (i.e., standardized mortality ratios), whereas those in the appendix used direct standardization in response to the external review. Since the specific cancers are rare, it is not clear how to interpret the comparison between results using indirectly standardized rates as in the body of the IEI report and directly standardized rates as in the IEI response to the external review. A further complication is that different time periods were used for the comparative analyses. In Chapter 2, the pre-PAVE PAWS period is 1960-1977, while in the Appendix the pre-PAVE PAWS period is 1969-1977. There are two post-PAVE PAWS periods in both analyses. In Chapter 2, the first comparison period is 1978-2001, while in the Appendix it is 1978-1989. Direct comparison of the 1978-2001 time period with the 1978-1989 time period is problematic. For the second comparison period, data in Chapter 2 are for 1990-2001, while in the Appendix they are for 1990-2002. While not identical, the 1990-2001 and 1990-2002 time periods can be compared with some confidence. There was no explanation or justification for how the time periods noted above were chosen, even though for small populations such as these the inclusion or omission of a few years can have a large effect on average rates and comparisons. Note the year-to-year variations in the cancer rates and the ratio of the data from Barnstable and the comparison counties in the pre-PAVE PAWS period in Figures 2.5, 2.6, 2.7, 2.8, and 2.9 in different years of the IEI report (IEI 2005). Despite these methodological issues, the committee conducted some comparisons using the available information and found some substantial differences. Without additional information, the committee is unable to determine whether the differences reflect an inappropriate comparison population, methodological differences, different time periods of observation, or other explanations. Therefore, the results presented in Chapter 2 cannot be fully 1 Standardization is a process whereby age specific rates (e.g., cancer rates for each five-year age group, 0-4, 5-9, 10-14, … 70-74, 75-80 …) are averaged using weights proportional to the relative size of the total population in each age group. Direct and indirect approaches use different weights to calculate these averages. In situations where there are small numbers of events, such as for rare cancers, these two methods can result in substantially different rate averages.
OCR for page 5
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar interpreted and the committee is reluctant to exclude the possibility of differences in mortality rates in the pre- and post-PAVE PAWS time periods. Other analyses are also difficult to interpret due to the specific methodological approaches used, such as the failure to use exposure data. The committee notes that IEI’s trend analysis does not incorporate exposure measurements but rather uses the date of first operation of PAVE PAWS as a proxy for possible exposure among those living in Barnstable County (i.e., anywhere on Cape Cod). Aggregating mortality across Barnstable County results in diluting the associated exposure to PAVE PAWS, since exposures across the county cover the entire dynamic range of exposure as indicated in Figures 1.2 and 5.A1 of the IEI report. The 2005 NRC report estimated that only 11.8 percent of the upper Cape Cod population (a subset of Barnstable County) was within line-of-sight for side lobes of PAVE PAWS, and exposures can vary by a factor of more than 8,000 for residents of this county. As a result, the relevance of the secular trend analysis is limited by including individuals experiencing very different exposures to PAVE PAWS within the Barnstable County group. The range of exposure values could result in either false negatives, due to the small proportion of high-exposure individuals, or false positives, perhaps due to confounders among the low-exposure individuals. Pre- and Post-PAVE PAWS Comparisons and Time Trends in Mortality The secular trend analysis of the IEI report explored specific aspects of secular trends in mortality but leaves some trend questions unanswered. The IEI secular trend comparison addresses the question: Does the change in indirectly standardized mortality rates before and after initiation of PAVE PAWS operations differ between Barnstable and the comparison counties? The IEI approach compares the changes in two specific ratios that represent the cumulative disease experience over each time interval. To further address the changes in temporal rates, the IEI report presents graphs of three-year moving average age-adjusted mortality rates, raising a more specific question: Does the annual change in age-adjusted mortality after PAVE PAWS differ between Barnstable and the comparison counties? This second question is not directly addressed by the secular trend analysis as described in the text of the IEI report, but this question is raised by the data presented in figures such as Figure 2.8 (male brain cancer), which suggests a diverging trend of increased brain cancers in males in Barnstable County relative to the comparison counties after the 1987-1989 time interval. However, the wide variability in three-year moving averages and the small number of cases prior to 1978 exhibited in such figures complicates quantification of apparent trend differences. This change in variability may be due in part to changes in the population of Cape Cod. These two questions focus on different aspects of temporal trend (pre- and post-PAVE PAWS changes and annual trends since PAVE PAWS began operation, respectively), suggesting the need for clarification of the IEI report’s interpretation of the ratio of standardized mortality ratios as a measure of the “extent to which the trend in mortality rates changed to a greater (or lesser) extent in Barnstable County than in the comparison counties” (p. 35). As noted in the 2005 NRC report, some of the challenges encountered in studying the possible adverse health effects from PAVE PAWS are the small number of people living on Cape Cod and the even smaller number of people exposed to radiofrequency energy from PAVE PAWS. Furthermore, from an epidemiological perspective, the occurrence of some types of adverse health effects, particularly some forms of cancer, is relatively rare. When analyzing the potential association between an exposure, such as PAVE PAWS radiofrequency energy, and a disease, the fewer the number of people in the study population and the rarer the health effect under study, the less likely it is for a study to show a statistically significant effect, even if a true effect exists.2 That is, these studies have relatively low statistical power. The only way to increase the power, or the likelihood of finding a statistically significant effect, is to increase the number of subjects. Because of this, epidemiologists often focus more on the size of a relative risk, or the effect measure. For example, a relative risk of 2 might be considered important and warrant further investigation even if it is not statistically significant and 2 Comments on power, effect size, statistical significance, and p values are noted in Appendix B of this report.
OCR for page 6
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar if the study population is small. One must be cautious in interpreting the results because relative risk elevations could also be due to random fluctuations. Another approach for increasing sensitivity is to look at the number of cases of disease that occur over many years, thus increasing the sample size. However, this makes sense only if the effects are thought to be the same from year to year. In the IEI report the authors reported many relative risks that were not statistically significant and concluded that “widespread adverse health effects from PAVE PAWS radar exposure are unlikely” (p. 133). In considering their results, the committee recalculated some of the relative risks for secular trends in female leukemia mortality using the data available in the IEI report’s Appendix to assess the possible effect of using an inappropriate comparison population. While these data are not directly comparable to those used in Chapter 2 of the report, the committee is concerned that it obtained relative risks that average 1.5 and that these values are somewhat larger than the female leukemia rates provided in the IEI report, which average around 1.2 and are not statistically significant. The committee could not calculate the confidence intervals or statistical significance for the rates it derived because adequate data were not provided. While the observed differences between the committee’s results and those provided in the IEI report may be due to the standardization methods used and the specific time periods of comparison rather than a true effect, the committee is concerned that the results were very sensitive to small changes in how the calculations were conducted. Given the possibility of a substantially higher relative risk than that given in the report, the committee is reluctant to totally dismiss the possibility of an association between the radar and female leukemia and suggests that this issue warrants further consideration by IEI. Similarly, the committee recalculated the relative risks for female brain cancer mortality using data from the IEI Appendix and found that they average 1.4, somewhat larger than those in Chapter 2 of the IEI report, which average 1.15. Because of the potential impact on the conclusions of the IEI report, these discrepancies in analysis are important. In summary, the committee concludes that IEI’s use of Worcester County as a comparison group is problematic and that the IEI report would benefit from a calculation of rate ratios similar to those provided in Chapter 2 of the IEI report, excluding Worcester County. A comparison of this result with the previously calculated values would better indicate the extent of the problem, if one in fact exists. Second, the committee questions the methodology used in the IEI secular trend analyses and the exclusion of exposure data because of the wide dynamic range of exposures within Barnstable County. This limits the significance of the comparison with the other counties. Third, the report would be strengthened by estimating the differences between the temporal trends, such as those shown in Figure 2.8, and quantifying their uncertainty. Finally, with respect to the possibility of differences in mortality rates in the pre- and post-PAVE PAWS time periods, an estimate of the power of the study in terms of the expected disease rates and correlations with exposures that could be identified would greatly strengthen this portion of the IEI report. IEI Report, Chapter 3: Dose-Response Analyses of Mortality and Hospitalization Rates The IEI report used census-block-based Broadcast Signal Laboratory (BSL)(BSL 2004a, 2004b, 2005) measures of radar exposure (Chapter 1, section C-3, of the IEI report) and data for 248 block groups (61 census tracts) to evaluate surrogate3 dose-response relationships to address whether for each disease the rate increases or decreases as measured radar exposure 3 While the IEI report indicates that IEI monitored “dose-response” for exposures to PAVE PAWS radar, in fact the institute used exposure as a surrogate for dose. Dose would have to consider the amount of radiation to which a person was exposed, taking into consideration indoor versus outdoor exposures, times when the person was not in the area of exposure, and other factors. For microwave radiofrequency radiation (in the range of 0.1 MHz to 6 GHz) the specific-absorption rate equals the dose rate and therefore the specific absorption (dose) equals the integral of the specific-absorption rate.
OCR for page 7
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar goes from low to high. Cancer mortality data for 1987-2002 were available, as were hospitalization data for both inpatient and outpatient procedures (both provided by the Massachusetts Department of Public Health). Duration of residence was not taken into account. Rather, ZIP code of residence at time of death or hospitalization was used as the indicator of location and a radar exposure level was assigned for that location. The IEI report concluded that “analyses of mortality rates by census tract and census block group4 provided little evidence of adverse health effects resulting from exposure to PAVE PAWS radar” (p. 74) and that “analyses of hospitalization rates by ZIP code also provided little evidence of adverse health effects” (p. 75). In addition to cancer, the investigators selected a set of nonmalignant diseases to study with respect to exposure. Those diseases include neurodegenerative diseases (Parkinson’s disease, Alzheimer’s disease, and amyotrophic lateral sclerosis) and autoimmune diseases (multiple sclerosis, autoimmune thyroiditis, and systemic lupus erythematosis), which may not demonstrate a dose-response to nonionizing radiation. No justification for the selection of diseases is provided in the report, and certain common autoimmune diseases, such as rheumatoid arthritis, were not included. Mortality was examined for the neurodegenerative diseases, and hospitalization data for inpatient and outpatient procedures were used to examine the prevalence of both neurodegenerative and autoimmune diseases. Since many of the nonmalignant diseases under study are chronic in nature, it is unclear how the hospitalization data reflect the actual disease prevalence in the community. Importantly, the hospital utilization data were limited to one count per person, but no further indication is given as to how much of the total burden of disease is captured. Furthermore, some of these diseases (e.g., amyotrophic lateral sclerosis) are relatively uncommon; consequently there may be insufficient power for the analyses performed to detect an effect of exposure. Analyses of PAVE PAWS radar exposures in relation to mortality were adjusted for age and sex as well as for median income, percentage of adults living below the poverty line, percentage of black residents, mean education level, and number of nursing home beds, using 2000 census data. In addition, to address concerns regarding chemical exposures from the Massachusetts Military Reservation, the IEI report includes the distance from the southern boundary of the reservation as a covariate in its models. The IEI report provides no elaboration of the appropriateness of the distance-based surrogate for possible chemical exposures or its adequacy in addressing potential confounding effects (e.g., descriptions of water sources or distribution systems, historical water quality reports, and so forth), thus the committee can not adequately assess the appropriateness of this exposure surrogate. Finally, the IEI report also includes an additional covariate for exposure to pesticides from cranberry bogs in the breast and brain cancer analyses. The committee has several concerns with the analyses of both nonmalignant diseases and cancer outcomes. While residential address and its associated exposure level may reflect exposures relevant for cancer mortality, it is not clear that residential address for mortality from nonmalignant diseases included in this evaluation necessarily reflects long-term exposures because elderly people may move to be near a family member in the latter stages of life. In addition to the duration and magnitude of exposure, a problem not addressed in the IEI study is the latency or delay time from the initiation of a disease to the time of diagnosis or other endpoints. Since the latency and rates of disease progression vary from cancer to cancer, these data are needed for analysis of how PAVE PAWS radiation affects the initiation, promotion, or progression of a given cancer. This issue reduces the statistical power and limits interpretation of the reported findings. In the IEI report several analyses are presented in Chapters 3 and 4 to evaluate survival time following diagnosis, particularly the use of data from 1990 onward. A major concern with this approach is that the underlying disease process and population changes may preclude meaningful interpretation. At a minimum, the difference in the underlying 4 Census blocks are the smallest geographic unit for which census data are available. Blocks aggregate to block groups, which aggregate to tracts, which in turn aggregate to counties. ZIP codes do not map directly to any of these geographic units.
OCR for page 8
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar disease incidence and survival patterns (i.e., numbers of individuals with cancer vs. number of deaths from cancer) limits interpretation of cancer mortality data. For example, many breast cancer patients who died after 1990 would have been diagnosed before 1978, and an unknown percentage may have been diagnosed before having moved to Cape Cod. This concern does not apply to lung cancer, which has a very short average survival time (six months). Finally, the overall increase in population on Cape Cod precludes an analysis of latency using date of event for classification. The Silent Spring Institute overcame this limitation in its study of breast cancer by obtaining data from individual women to classify them according to their length of residence on Cape Cod (McKelvey and others 2004). As recommended by the 2005 NRC report, individual-level data are required to evaluate latency and provide meaningful dose-response analyses. IEI Report, Chapter 4: Dose-Response Analyses of Cancer Incidence Rates Cancer incidence was also evaluated in relation to PAVE PAWS radar exposure. The analytic approach was similar to that described for mortality. The authors of the IEI study note that “analyses of cancer incidence rates by census block group provided little evidence of adverse health effects resulting from exposure to PAVE PAWS radar” (p. 103). This finding is consistent with the committee’s 2005 analysis of cancer incidence at the census-tract level (NRC 2005, p. 165). A low geocoding success was noted for one mid-Cape Cod ZIP code (51 percent) and one upper Cape Cod ZIP code (52 percent), and 28 census block groups within seven ZIP codes had geocoding rates below 60 percent. The IEI report provides general reasons for the low rates but does not specify the proportions of missing geocodes due to rural route/P.O. Box addresses or other reasons, nor does the report describe temporal patterns in the missing data. No explanation is provided for these low geocoding rates, and the differential among ZIP codes could result in bias. IEI Report, Chapter 5: Dose-Response Analyses of Birth Weight Chapter 5 reports the detailed analyses of birth weight in relation to PAVE PAWS average power-density measures. Analyses were based on births from 1989 to 2002 and addressed a wide range of potential confounders. Based on this evaluation, and in conjunction with previous work, it is most unlikely that PAVE PAWS radar exposure has had an adverse effect on birth weight. 2. REVIEW THE CONCLUSIONS REACHED IN THE 2005 NRC REPORT IN LIGHT OF THE IEI DRAFT FINAL REPORT The committee considered all conclusions in the 2005 NRC report and determined that the IEI report was only relevant to the epidemiological conclusions of the NRC report (NRC 2005, pp.180-181). The committee’s initial conclusion regarding health effects was that the analyses performed in its 2005 report did not indicate any increase in cancer risk with estimated peak or average power density exposure to the PAVE PAWS beam. This conclusion is not changed by the IEI report, in part due to the limitations of the data and the IEI analyses as noted above. 3. REVIEW THE NRC COMMITTEE’S RECOMMENDATIONS FOR FOLLOW-ON EPIDEMIOLOGICAL STUDIES The IEI results do not significantly alter the committee’s previous epidemiological recommendations that additional improvements in study protocols would provide more definitive
OCR for page 9
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar information regarding the potential for health effects (see Appendix C of this report for the NRC 2005 report epidemiological recommendations). Nevertheless, the committee notes that while further studies (such as individual dose-response measurements) would clarify some of the issues raised in the IEI report, it is not clear that such studies would result in evidence sufficient to indicate that there are or are not adverse health effects resulting from PAVE PAWS radar. In the 2005 NRC report (pp. 155 and 157), the committee noted that it is unclear what minimal level of risk would provide sufficient assurance that the PAVE PAWS facility is not adversely impacting health. If the answer is 99 percent assurance that there is not an increase of greater than 1 percent, then research would not be able to satisfactorily address such concerns. Using cancer as an example, and assuming 10 years of observation of the population, one could reasonably anticipate that well-designed and well-conducted research, either prospective or retrospective in design, could address moderately low levels of relative risk, such as between 1.4 and 2.0, (increases in risk levels between 40 percent and 100 percent) for some of the more common cancers (such as colo-rectal, lung, breast, prostate, bladder and non-Hodgkin’s lymphoma). The NRC committee also made specific recommendations regarding future epidemiological studies of the PAVE PAWS radar. On pages 181 and 182 of the 2005 NRC report, the committee recommended that “Because of the limitations of exposure estimates, confounders, and range of health outcomes it is recommended that future investigations integrate: (1) exposure assessment and health at the census block level; (2) personal exposure characteristics other than just residential location; (3) extensive consideration of potential confounders; and (4) health outcomes other than cancer. In addition, it is recommended that geographical correlation studies should be carried out for age-specific strata … (e.g., under 30 or 45)”. While the committee recognizes that the IEI study was ongoing at the time the committee’s recommendations were published, it comments here on our previous recommendations in relation to the draft IEI report. NRC Recommendation 1 The IEI report satisfies, in part, NRC Recommendation 1 (integrate exposure assessment and health at the census block level) by considering ZIP code areas, census tract, and census block group data, but stops short of census-block-level data, primarily due to limited data availability at the block level. For example, income level for socioeconomic status calculations is only available at the block group level, and hospitalization data were only available at the ZIP code level. A more complete treatment of Recommendation 1 would involve improving geocoding coverage to the block level and obtaining, at a minimum, block-level covariate information. Moving to smaller geographic areas, while recommended, would require the incorporation of random effects, to stabilize statistical estimates from small areas, with or without spatial correlation, as implemented in the 2005 NRC report. The available Broadcast Signal Lab measurement data and models of the PAVE PAWS power-density emissions provide a good first-order characterization of the spatial distribution of exposures occurring throughout the communities of Cape Cod. The committee’s 2005 report noted that despite the measurements and estimates of the PAVE PAWS waveforms (AFRL 2003) and power densities (BSL 2004b), there currently are no data for estimating personal exposure at the level of an individual. The strength of integrating exposure and health assessment at smaller geographic increments is that it provides a closer comparison of estimated local exposures and local incidence of health effects than does averaging heterogeneous exposures over a larger number of people to gain sample size (such as the increase in sample size associated with Cape-wide analysis comes at the loss of more accurate local exposure estimates). On the other hand, moving to smaller geographic areas results in smaller sample sizes, hence larger associated local variances. A common methodological solution that increases geographic resolution while maintaining statistical precision, is the use of random effects to stabilize statistical estimates from small populations by borrowing information from neighboring regions. Such methods are detailed on pages 154-157 of the 2005 NRC report.
OCR for page 10
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar NRC Recommendation 2 Regarding the committee’s second recommendation (integrate personal exposure characteristics other than just residential location), the IEI report used residential location (aggregated to ZIP code area, census tract, or block group level) to determine exposure values through population-weighted averages of the BSL power density measurements. The BSL measures are based on outdoor levels away from buildings. As stated in the 2005 NRC report, incorporating personal exposures remains important to addressing time and activity patterns of individuals and, at a minimum, to assessing the appropriateness of using the BSL values as exposure surrogates. A phased approach to implementing this recommendation is possible. For example, the first step would involve a comparison of BSL measured values to personal levels monitored via dosimeters carried by individuals over a period of time. Samples should be constructed so as to include a set of individuals covering the full range of BSL power-density values and the full range of local incidence or mortality rates and with adequate numbers to reliably describe personal exposures associated with appreciable changes in BSL values. Due to time spent indoors and in other shielded areas, personal exposures might reasonably be expected to be lower than ambient outdoor measured values, and such a study could provide insight into associations between the BSL measures and representative personal exposures (e.g., whether BSL values might be adjusted by activity patterns to better approximate personal exposures). Pending the outcomes of such studies, as a next step, case-control studies could construct and compare historical exposure values between diagnosed and healthy individuals. Subsequently, long-term prospective cohort studies could include personal exposure measurements and follow individuals’ health outcomes over an extended time. The IEI report did not address this deficiency. The strength of incorporating personal exposure measurements is that they would address time and activity profiles of individuals, provide information on the potential effects of shielding (such as time spent inside a home as opposed to outside), and provide information on the appropriateness of using BSL exposure estimates as surrogates for actual exposure measurements. The limitation is that personal monitoring would substantially increase the costs if future studies are contemplated. NRC Recommendation 3 Regarding NRC Recommendation 3 (integrate extensive consideration of potential confounders), the IEI report incorporates several standard confounding factors such as age and socioeconomic status and discusses potential impacts of smoking behavior in Cape Cod residents as related to patterns in cancer incidence and mortality (but without data relating to smoking, such as the Behavioral Risk Factor Surveillance System smoking patterns used in the 2005 NRC report). Additional confounders that would be of particular interest include regional variations in health screening availability and use, stage of diagnosis, and hospitalization rates. In the evaluation of a possible association between radiofrequency energy from the PAVE PAWS radar and health effects, exposures to other sources of RF energy, such as radio and TV transmitters or cell phones, are also of interest as potential epidemiologic confounders. Confounding factors can distort measured associations between outcomes and risk factors, so proper adjustment is needed to place results in the context of whether there is any presence or absence of a possible effect of PAVE PAWS exposure on the health outcomes of interest. Evaluation of confounders would have the obvious strength of distinguishing health effects that arise from other causes from those that might arise from the PAVE PAWS radar. A limitation of this recommendation is that it may not be possible to isolate the effects of the radar, if any, from other environmental impacts. NRC Recommendation 4 Regarding NRC Recommendation 4 (integrate health outcomes other than cancer), the IEI report includes analyses of low birth weight, neurodegenerative disorders, and autoimmune disorders. The IEI low birth weight analysis has the added strength of individual-level data for geocoding (but still maintains aggregate PAVE PAWS exposure surrogates).
OCR for page 11
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar Based on the results of the IEI report and the committee’s related comments, the committee retains recommendations 1 through 4 of the 2005 NRC report and adds the following new recommendations. New Recommendation 1 More careful consideration should be given to the selection of comparison counties, including, but not limited to, population density patterns across sub-county units. In particular, the committee recommends that Worcester County not be used for comparison. Excluding Worcester County from the comparison would eliminate a possible source of bias, adding credibility to the comparison of counties with and without exposure to the radar. New Recommendation 2 Additional known and potential confounders such as smoking, hospital admission rates, pesticide exposures, occupational exposures, cell phone use, and power densities from radio and TV stations should be considered in any future studies. The strengths and limitations of this recommendation have been noted previously in the 2005 NRC Recommendation 3 above. New Recommendation 3 Any future studies of temporal trends should incorporate annual changes in estimated disease rates rather than basic pre- and post-comparisons of rates as calculated by IEI. These changes could be incorporated into small-area analyses such as those reported in the 2005 NRC report. This methodology allows inclusion of area-specific PAVE PAWS exposure surrogates and could provide results more focused on PAVE PAWS than comparisons of rates before and after the facility began operation. A trend in annual changes will make more use of the available data than merely comparing two averages. A trend of increasing risk with longer exposure would also be plausible for some of the health conditions under study. Furthermore, the proposed measure can be summarized over a time period with a single statistical test that retains power to evaluate the association of interest. Importantly, this measure more directly addresses the issue of interest than does the mean value before and after the start of PAVE PAWS operation. A limitation is that this analysis would be more labor intensive than a pre- and post-evaluation comparison of rates. New Recommendation 4 Based on the strengths of the IEI birth weight study (large sample size, individual level outcomes, and block-group-level exposure data and analysis), the committee recommends that no further studies of low birth weight be done. SUMMARY AND CONCLUSIONS The NRC Committee to Review the Potential Health Effects from Exposure to the PAVE PAWS Radar has reviewed the IEI draft final report (September 9, 2005 version) entitled Descriptive Studies of Disease Occurrence and PAVE PAWS Radar. The committee has noted a number of corrections, criticisms, and suggestions for how the study could be improved. The committee concludes that the IEI report has provided new information relative to the assessment of potential health effects of the PAVE PAWS radar. For example, the IEI report considered census block group data and data at the ZIP code level and included analyses of low birth weight, neurodegenerative disorders, and autoimmune disorders. In particular, the individual level low-birth-weight analysis suggests that extensions of the present studies are unlikely to find an association with exposure to the PAVE PAWS radar. However, the sensitivity of other IEI analyses to methodological choices, such as the time intervals and comparison counties, makes interpretation of IEI’s other results unclear.
OCR for page 12
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar In summary, the committee makes no change in its conclusions or recommendations as set forth in the 2005 NRC report with regard to the possible health effects of the PAVE PAWS radar. The committee does, however, propose additional recommendations to address methodological deficiencies in the IEI report if it is deemed that any further epidemiological studies are to be undertaken. Sincerely, Frank E. Barnes, Chair Committee to Assess Potential Health Effects from Exposures to PAVE PAWS Low-level Phased-array Radiofrequency Energy APPENDIX A Factual Errors There are a several statements in the IEI report that the committee finds to be factually incorrect or incomplete. On page 4, the statement that “frequency is increased over the pulse length” is incomplete. Many of the PAVE PAWS pulses are at a fixed frequency, although this frequency can change from pulse to pulse. In one of the track modes, some of the pulses use frequency chirp where the frequency varies during the pulse transmission with a maximum bandwidth of 2 MHz. Also on page 4, the statement that “precursors may be produced by the transient characteristics of the waveforms, in particular the signal bandwidth (buildup times), generated by certain types of radar” is incomplete. A detailed discussion of the sources and characteristics of precursor formation is provided in the 2005 NRC report. Briefly, all of the calculations and measurements that show observable precursors occur for short-pulse, very wideband signals. Calculations show that for the PAVE PAWS 4-MHz bandwidth there are no measurable precursors. Residual calculated precursors are many orders of magnitude below the radar noise levels. On page 5, the statement that “EM wave energy is directly related to frequency” is misleading. For radiofrequency exposures, the energy, energy density, and power depend entirely upon what parameters are kept constant. Both power and energy density, in watts per square meter, are proportional to the magnitude of the electric field squared. On page 76, reference is made to the Toler study on the impact of 435-MHz radiofrequency on neurotransmitter and hormone levels in exposed rats. An unfounded conclusion is stated that the persistent drop in dopamine levels in exposed animals is induced by stress. This conclusion is then used to support the epidemiological finding of no association between radiofrequency exposure and Parkinson’s disease. Such a description of the Toler study and conclusion is overreaching and unjustified. The idea of a dose-response relationship as described throughout the IEI report is misleading. Dose is a precise concept that takes into consideration a person’s total exposure, including length of time in the field, indoor versus outdoor exposures, and other considerations. This study actually monitors a dose surrogate, which is exposure level assigned to the census tract, block group, block, and ZIP code. For microwave
OCR for page 13
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar radiofrequency radiation (in the range of 0.1 MHz to 6 GHz) the specific-absorption rate equals the dose rate and therefore the specific absorption (dose) equals the integral of the specific-absorption rate. In the IEI report, careful review of time periods needs to be conducted. Based on population trends and person-years, there appears to be a correction necessary in the dates noted for the column “1969-1977” (for example, see Tables 2.2, 2.27, 2.28, 2.29, 2.30, and 2.31). APPENDIX B Comments on Effect Measure, Confidence Interval, Probability Value, and Statistical Power The IEI draft final report chapters include statistical measures such as rate ratios, 95 percent confidence intervals, and probability values. Statistical power is an issue when considering the meaning of these measures. Some basic descriptions of the measures are presented in this appendix. The authors of the IEI report interpreted their results in terms of several statistical measures, but they were less clear in terms of the relative importance placed on these measures in drawing their conclusions. Perhaps the most important measure used to describe the degree to which an individual with a specified risk factor (or exposure) is more likely to contract a disease than an individual without the specified risk factor (assuming the two individuals are identical otherwise) is the effect measure. The effect measure can be expressed as an odds ratio, rate ratio, risk ratio, or relative risk. In addition to reporting the size of the effect measure, statistical methods are used to determine whether any differences observed in disease rates are more or less likely to be due to chance fluctuations and to communicate the amount of uncertainty in the findings. Effect Measure Each of the effect measures estimates the relative risk of disease in one populationfor example, those exposed to the highest measured average power-density level from the PAVE PAWS radar to another population, such as those exposed to the lowest measured average power density level. If the risks in the two populations are the same, the relative risk is 1.0, showing no difference or no effect. Table 4.1 (p. 106) of the IEI report presents the age-adjusted relative risk of male brain cancer incidence as 1.27 (or about 1.3). This means that among those experiencing the highest exposures the risk of male brain cancer was about 30 percent higher (1.3 times as high) compared to those experiencing the lowest exposure. This is considered a small relative risk but still suggests an excess risk of disease among those with higher exposures. Relative risks greater than 2.0 are of particular interest, corresponding to doubling of the risk, and relative risks greater that 5.0 are considered unusual and substantial. Confidence Interval Another measure used in the IEI report is the 95 percent confidence interval. For instance, the risk of age-adjusted male brain cancer (Table 4.1, p. 106; see also Figure 4.1, p. 108) from exposure to PAVE PAWS is reported to be 1.3, with a 95 percent confidence interval from 0.8 to 2.0. This means the best estimate given the data available is that the risk for the highly exposed group is 1.3 times higher than that for the lowest exposure group, but the risk could be as low as only 0.8 times or as high as 2.0 times that of the lowest exposed group. In this case the finding of a best estimate of an elevated risk of 1.3 is not considered to be statistically significant because the 95 percent confidence interval includes the outcome that there is no elevated risk (relative risk equals 1.0). Note that the size of the confidence interval is directly related to number of people in the study and the number of disease cases in the study
OCR for page 14
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar population. The more people in the study, the more accurate the estimate of disease risk and the smaller the confidence interval. Therefore, in studies with small numbers of subjects, confidence intervals are often very wide and may be less meaningful. Probability Value In addition to presenting 95 percent confidence intervals for the estimates of relative risk, IEI calculated the probability value. The p value (probability value) is another statistical measure used to quantify uncertainty about whether an outcome is due to chance or actually reflects a true difference between populations. Using a formula that considers the number of subjects and cases, the percent likelihood of a chance outcome, for example a 1 percent likelihood, is expressed as a p value of 0.01. Traditionally, a p value of 0.05 or less (5 percent or less) has been accepted as evidence of a statistically significant difference unlikely to be due to chance, which corresponds to a 95 percent confidence interval. As with confidence intervals, this value is very sensitive to the number of people and cases in the study and is less meaningful for small studies. Statistical Power Statistical power is the probability that an effect in an epidemiological study can be detected if one actually exists. A challenge of studying the possible adverse health effects from PAVE PAWS is that the number of people living on Cape Cod, particularly those in close proximity to and/or exposed to radiofrequency energy from PAVE PAWS, is small and most cancers are relatively rare, at least from an epidemiological perspective. The smaller the study population and the rarer the health effect under consideration, the less likely it is for a study to show a statistically significant effect, even if a true effect exists. The lack of statistical significance does not necessarily mean that there is no effect, but rather, given the small number of cases that the study design is unable to show to a high degree of probability (>95 percent) that the exposure is associated with disease. Therefore, discounting a risk factor because a study does not show a statistically significant effect may be misleading. It may be that the effect is simply smaller than can be shown with the number of people and cases in the study but is important nonetheless. Further details on statistical power are provided on pages 154-156 of the 2005 Committee report (NRC, 2005). APPENDIX C Recommendations from 2005 committee report entitled, An Assessment of Potential Health Effects from Exposure to PAVE PAWS Low-Level Phased-Array Radiofrequency Energy The following is excerpted from Chapter 10 (Summary of Conclusions and Recommendations) of the committee’s 2005 report. Because the committee judged that the epidemiology recommendations are the most relevant to the current letter report, it provides them immediately below followed by the biology recommendations. Recommendations for Epidemiology Studies Because of the limitations of exposure estimates, confounders, and range of health outcomes it is recommended that future investigations integrate: (1) exposure assessment and health at the census-block level; (2) personal exposure characteristics other than just residential location; (3) extensive consideration of potential confounders; and (4) health outcomes other
OCR for page 15
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar than cancer. In addition, it is recommended that geographical-correlation studies should be carried out for age-specific strata (e.g., under 30 or 45). Biology Recommendations Given the observation of long-term RF-exposure effects in plants and the several (though unreproduced) reports of effects in intermediate-duration exposures, the committee concludes certain additional studies on the biological effects of PAVE PAWS-specific RF exposures are warranted. In particular, studies using large-scale genomic screening for gene and protein expression at the cellular level and studies of plant growth in the area around the PAVE PAWS facility would improve our understanding of possible biological effects of long-term exposures to this radar. While biological responses do not necessarily translate into health effects, it is important to follow-up on known biological effects to determine if they relate to an identifiable health effect. The committee concludes that there is a need for three types of studies: Studies are recommended that include large-scale genomic and proteomic screening to identify gene- and protein-expression patterns in cell and animal studies after exposure to simulated PAVE PAWS radiation, both at levels approximating peak-exposure levels of the CAPE COD population and at higher power levels to identify potential threshold power densities for biological effects. The committee recognizes that the limitation of these studies is that they may have limited applicability to the human situation. However, the strength of these studies is that they will provide information covering a wide range of cellular activity in a limited number of experiments and the information gained may be useful in generating theoretical mechanism-related hypotheses of the possible effects of PAVE PAWS exposures. Studies to further investigate the potential influence of PAVE PAWS exposure on neurotransmitters (e.g., dopamine) concentrations in the central nervous system are also recommended. The observation of a robust depression in dopamine levels (~50 percent) with an onset concurrent with the start of exposure, and lasting for the duration of the exposure, could prove to be of great importance if this effect can be replicated. Dopamine is closely involved in motor control (e.g., depressed levels of dopamine are believed to be causal in the etiology of Parkinson’s disease) and so there could be a direct link between these observations and health effects. It is recommended that these studies be undertaken utilizing modulation frequencies more representative of the PAVE PAWS system rather than the 1 KHz utilized by Toler and others, as these lower modulation frequencies are closer to dominant brainwave frequencies. Studies of tree growth in the vicinity of the PAVE PAWS facility are also recommended. Evaluation of tree-ring width comparing rings from trees before and after the facility became operational with similar trees in areas that provide similar growth conditions outside the beam, should be possible with minimal or no impact on the environment. While these studies have the limitation of not being directly applicable to human health, they have the specific strength of involving long-term exposures (years). In addition, if the findings replicate the Latvian tree-growth results, it is anticipated that the information from these studies could lead to mechanism-generating hypotheses. REFERENCES AFRL (Air Force Research Laboratory). 2003. Phase IV—Time-Domain Waveform Characterization Measurements of the PAVE PAWS Radar. Kirtland Air Force Base, Albuquerque, NM.
OCR for page 16
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar BSL (Broadcast Signal Laboratory). 2004a. Survey of RF Energy Field Emissions from the PAVE PAWS Radar Located at Cape Cod Air Force Station, Massachusetts: Final Test Plan, prepared in consultation with the International Epidemiology Institute, Rockville, MD, by Broadcast Signal Lab, Medfield, MA. BSL (Broadcast Signal Laboratory). 2004b. A Survey of Radiofrequency Energy Field Emissions from the Cape Cod Air Force Station PAVE PAWS Radar Facility: Final Test Report, prepared for PAVE PAWS Public Health Steering Group by Broadcast Signal Lab, Medfield MA. IEI (International Epidemiology Institute). 2005. A Public Health Evaluation of Radiofrequency Energy from the PAVE PAWS Radar, Cape Cod Air Station, Massachusetts: Draft Final Report. Descriptive Studies of Disease Occurrence and PAVE PAWS Radar. McKelvey, W., J. G. Brody, A. Aschengrau, and C. H. Swartz. 2004. Association between residence on Cape Cod, Massachusetts, and breast cancer. Ann Epidemiol 14:89-94. NHS (National Health Service). 2003. Strang J. R. A Descriptive Study of Cancer Occurrence Around RAF Fylingdales, 1991-2000, Scarborough, Whitby and Ryedale Primary Care Trust, NHS, United Kingdom. NRC (National Research Council). 1979. Analysis of the Exposure Levels and Potential Biologic Effects of the PAVE PAWS Radar System. Washington, DC: National Academy Press. NRC (National Research Council). 2002a. Letter Report to the Department of the Air Force from the Committee to Assess Potential Health Effects from Exposure to PAVE PAWS Low-Level Phased-Array Radiofrequency Energy: Recommendations for Phase IV Measurements. Available online at http://www.nap.edu/books/NI000468/html/. NRC (National Research Council). 2002b. Interim Letter Report to the Department of the Air Force from the Committee to Assess Potential Health Effects from Exposure to PAVE PAWS Low-level Phased-array Radiofrequency Energy: Adequacy of Available Research Research Data. Available online at http://www.nap.edu/books/NI000483/html/1.html#pagetop. NRC (National Research Council). 2005. An Assessment of Potential Health Effects from Exposure to PAVE PAWS Low-Level Phased-Array Radiofrequency Energy. Washington, DC: National Academies Press.
OCR for page 17
Letter Report to Review the International Epidemiology’s Public Health Institute’s Evaluation of Radiofrequency Energy from the PAVE PAWS Radar COMMITTEE TO ASSESS POTENTIAL HEALTH EFFECTS FROM EXPOSURES TO PAVE PAWS LOW-LEVEL PHASED ARRAY RADIOFREQUENCY ENERGY FRANK E. BARNES (Chairman), University of Colorado at Boulder, Boulder, CO ROBERT C. HANSEN (Vice-Chairman), R.C. Hansen, Inc., Tarzana, CA LARRY E. ANDERSON, Battelle Pacific Northwest National Laboratories, Richland, WA GRAHAM A. COLDITZ, Harvard Medical School, Boston, MA KATHLEEN A. COONEY, University of Michigan, Ann Arbor, MI FRANCESCA DOMINICI, Johns Hopkins University, Baltimore, MD KENNETH J. McLEOD, State University of New York, Binghamton, NY KEITH D. PAULSEN, Dartmouth College, Hanover, NH SUSAN L. SANTOS, University of Medicine and Dentistry of New Jersey and East Orange, War Related Illness and Injury Study Center, Medford, MA JAN A. J. STOLWIJK, Yale University, New Haven, CT LANCE WALLER, Rollins School of Public Health, Emory University, Atlanta, GA DANIEL WARTENBERG, Environmental and Occupational Health Sciences Institute, Robert Wood Johnson Medical School, Piscataway, NJ GAYLE E. WOLOSCHAK, Northwestern University, Chicago, IL National Research Council Staff RICK JOSTES, Study Director EVAN DOUPLE, Scholar MARILI ULLOA, Senior Program Assistant
Representative terms from entire chapter: