PUBLIC HEALTH ASSESSMENT

Community Exposures to the 1965 and 1970 Accidental Tritium Releases

LAWRENCE LIVERMORE NATIONAL LABORATORY, MAIN SITE (USDOE)
LIVERMORE, ALAMEDA COUNTY, CALIFORNIA

APPENDIX 1: COMMENTS ON THE PUBLIC RELEASE VERSION OF PHA (Cont.)

Reviewer 4 Comments

127. Further review of this document should be undertaken to improve its clarity and scientific credibility. It would be very helpful if both a conservative, deterministic calculation of dose for the most sensitive, maximally exposed individual(s) and a "best estimate" of a realistically expected dose were computed for purposes of comparison to the Monte Carlo analyses. The doses calculated in the present report do not represent realistic distributions, as they should do if one is to say with certainty that only 5% of the time will the dose be exceeded. Given the many very conservative assumptions used for the deterministic calculations, which affect the stochastic calculations, the chance of the reported dose being exceeded must be much less than 5%. Furthermore, the reported means must be higher than any realistic best estimate. The purpose of these calculations should be well explained; the equations and their applications must be transparent.

ATSDR Response: ATSDR has submitted this document to independent external peer review. Comments from all of the reviewers have elicited considerable revision to the document. The revised document also includes a reference to a conservative, deterministic dose estimation from the RASCAL model.

128. Understandably, this kind of material is very difficult to write-up clearly, especially when the target audience includes an extremely wide range of backgrounds. Overall, some improvement in understandability would be gained by some structural changes. In particular, the description of the assembly of the Monte-Carlo modeling process would be easier to follow if the various parts were first given an overview in the order in which the processes they model occur in the 'real world', followed by the current more detailed discussions of each process.

ATSDR Response: As above, the document has been extensively re-written. Specifically, the introductory portion of "Section 3. Exposure Assessment and Dose Estimation" and the appendix on the description and assumptions underlying the Monte Carlo analysis have been revised.

129. The old misinformation on the 1965 release must be carefully replaced with the new information. When the corrected information about the 1965 release is introduced, extreme care must be taken to avoid confusion between the 1965 and 1970 locations and distances. As the report now stands, the reader gets very confused between the similar, yet different, assumptions for the two years. A table may help make the differences clear.

ATSDR Response: The revised calculations for the 1965 release are based on newly available weather data. Figure 1 has been revised to clearly indicate differences in plume footprints and potentially exposed populations. Consequently, the section on the 1965 release has been completely re-written on the basis of the revised weather data and dispersion calculations.

130. HT converts extremely rapidly to HTO when it comes in contact with the soil. Unconverted HT diffuses out of the soil when the concentration in air becomes less than the concentration in soil. Deposition velocities are calculated using the measured concentration of HT in air and the measured concentration of HTO in surface soil. Thus soil tritium concentrations calculated using the deposition velocity refer to HTO only. HTO is then emitted to the atmosphere over time.

The assumption of oxidation of HT to HTO occurring just prior to emission is not based on any experimental evidence. Thus, all soil tritium should be referred to as HTO. It's conceivable that some HT is trapped in the soil and gets converted later, rather than instantly, but this fraction is too small to be measured.

Nowhere in any of Brown's papers does he talk about "transformation", although he does talk about oxidation, which is what seems to be meant by transformation. The assumption that HT is converted to HTO just prior to emission of HTO will not result in a different estimate of air concentration than the assumption that all tritium in the soil has been HTO from the time of HT deposition. However, the assumption of gradual transformation contradicts current knowledge.

ATSDR Response: The explanation of HT oxidation and subsequent emission of HTO to the atmosphere or uptake by plants has been changed as suggested.

131. The reader needs to have the assumptions summarized in a much clearer manner. What is presented is a collection of very conservative assumptions coupled with stochastic distributions that may or may not have a conservative bias.

ATSDR Response: The revised introduction states that basis for the public health decisions developed from these dose estimates is an evaluation of doses to the maximally exposed individuals. While we are trying to accurately convey the uncertainty in the dose calculations, ultimately we must ensure that the predicted doses do not underestimate actual historic doses. We have re-written Appendix 4, which explains the Monte Carlo simulation and parameter assumptions.

132. On each page where typing occurs at the very top, the first two sentences are indented by one space.

ATSDR Response: After several frustrating hours trying to determine why the word processing program is inserting those spaces, we have determined that they don't really matter.

133. p. iii, 2^nd paragraph: If "there are insufficient historic environmental sample data available to adequately evaluate past exposures", the rationale for the use of measured concentrations in vegetation and milk to calculate your ingestion doses for 1970 should be stated.

ATSDR Response: This paragraph has been changed to indicate that historic data are not sufficient to evaluate total tritium doses and that measured data are incorporated into the evaluation.

134. p. iii, 3^rd paragraph: The 1965 release occurred on January 20, 1965, not the 21^st.

ATSDR Response: All references to the 1965 release have been corrected to the January 20 date.

135. p. iii, 4^th paragraph, line 5: "In air" should be added so that the sentence reads"insufficient environmental measurements of each of the tritium forms in air following.."

ATSDR Response: This sentence has been changed as suggested.

136. p. iv, last paragraph: The ATSDR 2002 expert panel report conclusion that the annual contribution from chronic releases was never a public health hazard should be referenced.

ATSDR Response: The paragraph indicates that short term plus annual or chronic doses are not a public health hazard. Additional reference to the long term doses estimated by the review panel is included in Section 3.

137. p. 4, Figure 1: The caption mentions both accidents, but the one for 1965 is not shown (wind to east northeast). Yet the demographic data presented is for 1965, not 1970 as printed.

ATSDR Response: The revised figure shows different areas of exposure for the 1965 and 1970 releases as well as separate population estimates for each release.

138. p. 5, 2^nd paragraph, last sentence: Rephrase. The entire plume area covered much more than one census tract. Perhaps what is meant is that the census tract included the entire modeled plume area, which accounts for the highest concentrations and dose to receptor.

ATSDR Response: This sentence has been changed as suggested.

139. p. 5, 3^rd paragraph: "number of buildings"--Was there any other restriction on size or type of buildings, e.g., were outhouses counted? A description of the process may be appropriate.

ATSDR Response: Any building visible in the aerial photographs was counted. The text and footnote explain that this is a conservative estimate of total, potentially affected population.

140. p. 5, last paragraph: Nine people from the 1965 plume? This is the same number as the earlier draft, and it may (should) have changed with the new demographics. Is it reasonable to assume that the population density of the plume area is less than the tract average density? This might not be the case if the tract has huge unpopulated areas.

ATSDR Response: In response to newly available weather data for the 1965 release, both the location of the dispersed plume and the potentially affected population have changed. The document has been revised to reflect those changes. The area-proportional method of estimating the population of a portion of a census tract assumes that the population is uniformly distributed. For large, sparsely populated tracts, such as 4511, a reasonable assumption would be that the population density is higher in areas close to the urban centers and decreases with distance away from urban areas.

141. p. 7, title: In general, "stakeholder" concerns would be more accurate than "community" concerns.

ATSDR Response: Stakeholders are members of the community.

142. p. 8, 1^st paragraph: Figure 3 is called out here before Figure 2 is mentioned, but it sounds as if Figure 2 is meant.

ATSDR Response: Figure 2 is referenced on page 3. The page 8 reference correctly identifies Figure 3.

143. Fig. 3 is introduced in paragraph 1 as a "conceptual diagram" (only) but when looking at it one discovers it is more than just conceptual

ATSDR Response: The term "conceptual" has been removed.

144. Having said earlier in paragraph 1 "a significant portionis re-emitted", a qualifier should be added to the phrase, "taken up by vegetation", i.e., "some" is taken up, or "a little is taken up", etc. Last sentence: The "or" is misleading. It could be changed to "and". The vast majority of OBT is formed when HTO enters the plant as vapor through the stomata and is photosynthesized.

ATSDR Response: We believe the sentence is both clear and technically correct as stated.

144. p. 8, last paragraph: Shouldn't "flow path illustrated in Figure 2" really reference Figure 3? Shouldn't "human exposure and uptake as illustrated in Figure 3" really reference Figure 2?

ATSDR Response: These changes have been made as suggested.

145. Last phrase: "factors represent sources of uncertainty"--not really: Uncertainty about values to use for these factors causes uncertainty in the exposure estimate.

ATSDR Response: The phrase "sources of uncertainty" has been replaced with "variables".

146. p. 9, Figure 3. What does "The emission rate declines as the HT is transformed to HTO" mean?

ATSDR Response: This phrase has been revised to "The emission rate decreases over time as HTO is lost from the soil." Additional discussion and references about the time dependence of the HTO emission rate have been added to the text.

147. p. 10, 1^st paragraph: The phrase "hypothetical maximally exposed individual" could be a hot button phrase for some readers. The maximally exposed individual is a real person, somebody who was in the wrong place at the wrong time. He/she is not hypothetical. In other words, avoid modeling jargon here.

ATSDR Response: The definition of the maximally exposed individual has been expanded and includes the rationale for describing this individual as "hypothetical."

148. p. 10, 2^nd paragraph: The word "cumulative" is used without any context explanation, or context that could indicate what it means.

ATSDR Response: The word "cumulative" has been replaced with "estimated total."

149. p. 10, last two paragraphs: Apparently, the purpose of these is introductory, or as an overview of the following sections. But it's not entirely clear this is the purpose, and some of their content seems too detailed for an intro/overview. The purpose should be stated clearly.

ATSDR Response: These two paragraphs were redundant with following sections and have been deleted.

150. p. 11, 1^st paragraph: Reference the expert panel report.

ATSDR Response: There is no reference to information from the expert panel report in this paragraph. The phrase "for the RASCAL model" has been added to the second sentence.

151. Some explanation of why a Monte-Carlo procedure solves the HTO (RASCAL models HTO only) vs HT (the release was HT, not HTO) problem is needed. It would be helpful to follow the statement that RASCAL cannot calculate doses for HT by a statement that it can, however, estimate HT concentrations. From which this report then estimates doses by other methods (i.e., not RASCAL). Otherwise the reader might be wondering, well, if it can't do HT, why are you using it?

ATSDR Response: This paragraph has been revised to indicate why RASCAL is useful for estimating HT dispersion but not for evaluating the resulting HTO dispersion or concentrations. We have also included a comparison of RASCAL doses with those estimated using the Monte Carlo simulation in a subsequent section.

152. You refer to input and summary information in Appendix 1, but in Appendix 1, model outputs precede model inputs. It would be clearer if input information preceded output information.

ATSDR Response: Summary information (Tables A-4 and A-5) have been re-ordered as Tables A-1 and A-2.

153. p. 11, 3^rd paragraph: You need an explanation why 30 minutes was used when the previous paragraph refers to a 60-minute interval.

ATSDR Response: The 60 minute interval is a time range during which the release was assumed to occur. Newly available information about the release indicates that it occurred at 5:49 AM. The release is still assumed to have occurred over a 30 minute period, but the time references have been corrected to the beginning of release at 5:49 AM.

154. p. 11, 4^th paragraph: The reference is Brown et al (1990). Furthermore, the range of deposition velocities quoted is between 0.00041 m/s and 0.0013 m/s.

ATSDR Response: The references relating to deposition velocity have been corrected and expanded. As you suggest, the Brown et al. (1990) paper is a better reference for these velocities. With regard to the specific range of values, the velocity distribution we have assumed uses these upper and lower limits as 10^th and 90^th percentiles which includes a broader range of values and is consistent with the expanded list of deposition velocity references.

155. The units in which "deposition rate" is expressed need explanation. "m/sec" suggests to a reader without modeling expertise that a layer 1 meter thick is deposited in 1 second. That is "deposition rate" = "how much deposited in a unit of time."

ATSDR Response: The term "deposition rate" has been replaced with the term "deposition velocity".

156. A brief statement of why the normal distribution was selected might be helpful. It might be based on material in Brown et al., i.e., perhaps some indication in that paper of what they thought was likely/unlikely. A uniform distribution would better represent: "It is in this interval but we have no idea where in this interval, and no reason to prefer one part of the interval over another". The normal distribution does assert a belief that the middle is more likely than the ends. On the other hand, a uniform has fixed bounds, i.e. asserts it is not above or below certain values, whereas the normal has no such fixed limits (albeit, low probabilities for the extremes).

ATSDR Response: The distribution we have assumed uses the range of rates included in these references. The use of a uniform distribution requires knowledge of the upper and lower boundary values. Deposition velocities have been found to vary with soil type, soil moisture, vegetation cover, and season. Considering the range of parameters that affect deposition velocities, it is unlikely that the measured velocities capture the entire range of variation. Likewise, when integrating the range of velocities over a plume footprint, it seems reasonable to assume that the velocities will approach an average value. The normal distribution assumes that values higher than the mean are just as likely as values lower than the mean.

157. p. 11, footnote 7: RASCAL's output is apparently c/Q (s/m³), which is then multiplied by the release rate to give HT concentrations in air. This could be expressed more clearly. Furthermore, the concentrations referred to are not in Table 1 but rather in Table A-1.

ATSDR Response: This footnote has been re-written as suggested.

158. p. 12, 1^st paragraph. What is said in Brown et al (1988) is that the loss rate of HTO from soil over two weeks was 1% per hour or less.

ATSDR Response: We have expanded the discussion and references of the HTO loss rates from soil in the revised document. Further, in the revised calculations, the HTO loss rate has been modeled as an exponential decay probability distribution that varies from 0 to 8%/hour with an average rate of 1%/hour. Also, because the additional references indicate that HTO loss rates vary on a diurnal basis and are essentially zero during night-time hours, the revised ISC model uses a daytime emission factor of 1 and a night-time emission factor of 0. Consequently, the inhalation doses are based on the highest 12 hour averages rather than 1 hour averages.

159. p. 12, 2^nd and 3^rd paragraphs: Which Monte-Carlo analysis? According to the previous page there is already a Monte-Carlo for the deposition rate (although the term Monte-Carlo wasn't used there). Now there is one for the HT concentrations also? It would be easier to follow if the fact that the RASCAL point estimates are being replaced by a distribution of values was mentioned after paragraph 3 on page 11.

ATSDR Response: This section has been re-written. A statement explaining that point value RASCAL estimates have been replaced with assumed normal probability distributions has been added in a footnote and further clarified in Appendices 2 and 4.

160. p. 12, 2^nd paragraph, last sentence: Surely the estimated concentrations depend on distance from the source, so only one particular location on-site would be 3X higher than one location off-site (locations being distances along a centerline from the source).

ATSDR Response: The sentence in question states "The estimated maximum on-site concentration is approximately 3 times larger than the maximally exposed off-site location (1 mile from source)." This refers to only one on-site location and one off-site location. Using the revised atmospheric stability category, this statement is no longer true for the 1970 release. The highest concentration area corresponds with the 1 mile off-site location.

161. p. 12, 3^rd paragraph: Is there confusion between line one ("instantaneous maximum HT concentration") and "the mean value is the best estimate of the true value" (lines 5 and 6)?

ATSDR Response: The term "best" has been replaced with "most frequent".

162. p. 12, 4^th paragraph: Much of this is redundant - see 4^th paragraph on page 11.

Again, Brown et al (1990) mentions deposition velocities between 0.00041 m/s and 0.0013 m/s. Ogram (not Brown) et al (1988) gave a range of deposition velocities to field from 0.00027 - 0.0011 m/s. Deposition velocities to forest soil were given as 0.00033 to 0.0012 m/s. Also, it should be 1.82E-03 Ci/m³, not Ci-sec/m³.

A reason for selecting the normal distribution for the deposition rate should be provided. After all, the uniform distribution for the rate of HT to HTO conversion by soil microorganisms was justified on the basis of conservatism.

ATSDR Response: As previously stated, this section has been re-written with additional references and discussion.

163. p. 12, last paragraph: It should be noted which curve is on the top and which is on the bottom in Figure 4.

ATSDR Response: The curves and associated scales were color and symbol coded. However, this figure has been replaced.

164. p. 13, 2^nd paragraph: Why should the emission rate decrease just because total deposition decreases with distance from the source? Why do the rectangles cover between 10 and 30, while the sector identified in Figure 1 and Appendix 1 covers 10 to 50?

ATSDR Response: The HTO loss rate is the same for each area (%/hr); the HTO emission rate Ci/sec-m²) changes due to different cumulative air concentrations that vary by location. The revised document uses the terms HTO loss rate and HTO emission rate to clarify this distinction in terms.

The 1970 release covered the area between 10 and 30º. Due to greater dispersion, the 1965 release was assumed to cover the area from 10 to 50º. However, use of the newly available weather data indicates that the 1965 plume covered the area from 50 to 70º and does not overlap with the 1970 release. These changes have been made in the revised text and figures.

165. p. 13, last paragraph: Do you mean the ratio HT in air / HTO in atmospheric and plant moisture?

ATSDR Response: This sentence has been changed to indicate that "the HTO concentrations in soil, atmospheric, and plant moisture were in proportional equilibrium after a lag phase of 12-24 hours."

166. p. 14, Figure 4 should show the tritium in the soil as HTO. The legend should show the units for soil activity, and, for air activity, it should be Ci/m³ or Ci m^-3 but not a combination, as is written (Ci/m^-3).

ATSDR Response: This figure has been replaced (EXCEL does not like superscripted text).

167. p. 15, 2^nd paragraph and Table 1: Your lower deposition velocity (3.0E-4) does not agree with your lower deposition velocity mentioned on pages 11 and 12.

ATSDR Response: References to the lower deposition velocity have been corrected to the value of 0.0003 m/sec.

168. p. 15, 4^th paragraph, last sentence. Why would you say the air concentrations for 1970 and 1965 are equal, and where are the "following dose estimates"?

ATSDR Response: That statement refers to estimates of concentrations from the ISC models; the "following dose estimates" refers to subsequent sections wherein doses are calculated. On the basis of newly available weather data for the 1965 release and a revision of the atmospheric stability for the 1970 release, this section has been re-written.

169. p. 17, 1^st paragraph: You mean adult dose? Also, explain that the site boundary is 0.7 mile or whatever from the source. Also, it's very confusing to see this discussion of a comparison between the inhalation dose at the site boundary and the 12-day HTO dose, when Table 2 refers only to the 1 and 2-mile adult and child doses for inhalation HT exposure.

ATSDR Response: The referenced dose is for a child. The footnote on page 17 explains that the 30 minute doses were based on exposure at the site boundary. As you indicate, the explanation of locations for estimating the doses was unclear and has been revised. Use of the revised weather data has obviated the need for clarification. The locations of highest HT and HTO air concentrations all occur at the 1 mile location.

170. p. 17, Table 2: The results should be presented consistently in two significant figures. Also, there's confusion here with p. 22, paragraph 2, where you say inhalation dose from HT was estimated at the site boundary, which is less than 1 mile away.

ATSDR Response: As previously indicated, this section has been re-written. The rationale for estimating the 30 minute exposure from the site boundary was that a resident living at the 1 mile location could have been present for a 30 minute period at the site boundary during the initial release, but the longer, 12 day, HTO exposure would only occur at the residence location.

171. p. 18: The discussion here of ingestion exposure pathway is restricted to ingestion of food crops and dairy products, yet on page 10, paragraph 2, it states that the fourth component of the dose assessment is to estimate potential doses from food and soil. Accordingly, it should be clearly stated if childhood instances of "pica," for example, are also being addressed for purposes of the calculations. Additionally, it appears that the ingestion dose is based on U.S. data and not that for Livermore.

ATSDR Response: This assessment only includes ingestion from food items. The phrase "and soil" has been deleted from page 10. Relative to the inhalation and food ingestion components of dose, potential intake of tritiated soil moisture from incidental ingestion of soil particles is insignificant.

172. p. 19: The selection of normal distributions for the parameters mentioned should be explained.

ATSDR Response: A footnote explaining the basis for assuming normal distributions of HT deposition to soil and HTO concentrations in plants has been added to this section.

173. p. 19, 4^th paragraph: This is confusing. You seem to be reporting the differences between dose conversion coefficients for OBT shown in Table 2.5 of the ATSDR Expert Panel report. However, the implication of how the contribution of OBT is included in your dose estimates is extreme. It sounds as if the HTO dose (call it 1) is, multiplied by 2.2 (average; call it 2.2), and then added to the HTO dose (to get a dose of 3.2). The report of the expert panel (p. 7) says including OBT is "unlikely to produce a dose more than twice than of HTO alone."

ATSDR Response: The references to differences in dose conversion coefficients (from Table 2.5 of the Expert Panel Report) has been replaced by references to "percentage increase in dose from tritium as OBT above that from tritium in the form of water (HTO)" (from Table 2.3 of the Expert Panel Report). The percentage increase values in that table range from 16% to 400% and the panel concludes that the most likely range of increase is less than 32%. Consequently, we have assumed that the ingestion dose due to OBT is a triangular distribution ranging from 16% to 400% of the HTO dose with a most likely value of 32%. The text, tables, and figures have been revised as necessary.

174. p. 19, paragraph 5; and p. 20, Figure 6. Based on the geometry of the distribution in Figure 6, it is probably not appropriate to report in paragraph 5 on p. 19 that the "most probable dose is about 0.44 mrem." Instead, it is probably more correct to cite the 0.44 mrem dose average as the "expected value" or mean annual dose. p. 20, paragraph 2. It is probably more correct to identify the "most probable doses to children" as the mean or expected value doses. See comment directly above.

ATSDR Response: References to estimated doses have been revised to specifically indicate mean or average doses. The calculated distribution of doses indicates that most probable doses are lower than mean values and has been so noted in the discussion of total tritium doses.

175. p. 21, 1^st paragraph: You say "No limit was given for tritium" in DOE Order 5400.5. This is not true. Only radon 220 and radon 222 and their decay products are subject to DOE limits (II, 1, b); which may be different from the 10 mrem dose to the public for all other nuclides (II, 1, b, (1)). The 10 mrem/y standard expressed in 40 CFR 61 Subpart H (see DOE Order 5400.5 II, 1, b, (1)) most certainly applies to tritium releases, and in fact the monitoring and modeling of tritium emissions from LLNL has been a major part of NESHAPs compliance activities since their inception.

ATSDR Response: DOE Order 5400.5 (dated February 8, 1990) does list limits for tritium (water) and elemental tritium. There is no limit given for OBT. We will correct the text as required.

176. p. 21, Table 3: Needs two significant figures. Also, mention that the HT dose was calculated for the site boundary, while the HTO doses were calculated where? Sums for total tritium dose should equate properly between mrem and mSv, for example, 136 mrem is equal to 1.36 mSv (or if 1.35 mSv is derived, then the equivalent is 135 mrem. The adult 95^th-%tile summation for total tritium dose is 50.4 mrem not 51.3 mrem.

ATSDR Response: Table 3 has been revised as suggested (and in response to revised dispersion calculations).

177. The annual or chronic doses shown are those estimated by the expert panel for 1999. Conservatively, annual releases for 1965 and 1970 were perhaps 10 to 20 times higher than for 1999, and the estimated doses representing 1965 and 1970 would be correspondingly higher. The reader probably expects the chronic dose shown in the table to represent the chronic dose received by the same person who received the dose shown for the accidental release.

ATSDR Response: The annual adult doses applied to the revised total tritium dose are assumed to be 12 times the 1999 annual doses of 0.12 mrem per the Expert Panel's estimated values for historic annual doses (1.4 mrem/year). We have further multiplied the adult doses by a factor of 3 in estimating doses to a child (4.3 mrem/year). As the Expert Panel indicates that 1.2 mrem/year is a conservative estimate for the maximally exposed individual (adult), we have not presumed a 95^th percentile value.

178. p. 22, 3^rd paragraph: This says the HT doses were estimated at the site boundary. If they weren't, it should be changed.

ATSDR Response: The previous HT doses were estimated at the site boundary. However, as the revised dispersion calculations indicate that the maximum HT concentrations occurred at the 1 mile location, the revised HT dose estimates are based on the 1 mile location.

179. p. 23, paragraph 4. A comment should be made concerning the significance between a dose-response relationship for mental retardation that appears to have a threshold, and a dose-response relationship for impacts on intelligence and school performance that is linear, without a threshold.

ATSDR Response: Although both health effects relate to central nervous system development, the referenced study concludes that the collective data may support either a threshold or no threshold dose response. More important, is that, overall, these health effects occurred at doses that are 500 to a 1000 times higher than the doses estimated from the LLNL tritium releases. A statement clarifying the uncertainty of the dose response has been added.

180. p. 24, paragraphs 2 and 3. In utero fetal doses of 1 to 4 rad = 1 to 4 rem = 1,000 mrem to 4,000 mrem, and dividing each by 135 mrem makes the larger doses from 1 to about 30 times greater (not 1 to 10 times as noted on the next to last line of paragraph 2). Further, a dose level of 200,000 mrem total is greater than 135 mrem by a factor of more than 1,400 (i.e., 1,480), and not 140 (as noted at the end of paragraph 3).

ATSDR Response: According to a study commissioned by the Nuclear Regulatory Commission (Sikov et al. 1993) the fetal dose is approximately 1.7 times the dose to the mother (adult dose). A child dose is not an appropriate comparison. Potential fetal doses from the 1965 and 1970 releases and comparisons of those doses to relevant health studies have been revised accordingly.

181. p. 25 1^st paragraph: The expert panel estimated the dose from the 1970 release based on scaling with doses from chronic releases. This assumes the air concentrations are proportional to release rate, which is probably not true when an accident is compared with a chronic release. Although the bottom line may not change, this should be checked. Furthermore, the expert panel was talking about the risk from oocyte-labeling at oogenesis, and not neonatal (newborn) risk. Oocytes are mentioned in paragraph 4 on this page.

ATSDR Response: The Expert Panel made the assumption that 1% of the HT released from the 1970 accident was converted or present as HTO and calculated the resulting ooycyte labeling tritium dose accordingly (page 81; ATSDR 2002). They also recognized that chronic releases from LLNL were usually in the HTO form. Considering that the significant exposures to the HT plumes occurred after oxidation to HTO and the overall dispersion of the HT and HTO concentrations, the 1% conversion is reasonable. The phrase "early neonatal life" in the first sentence has been changed to "early prenatal life".

182. p. 25, paragraphs 3 and 4. The maximum annual total dose estimate in Table 3 on p. 21 of 135 mrem (for children) is almost 40 to >70 (i.e., 37 to 74) times lower than the 50,000 to 100,000 mrem total doses producing temporary low sperm counts, and not the 30 times noted. Moreover, Table 3 identifies maximum doses (mrem or mSv) and not administered activity concentrations, thus the comparison made in paragraph 4 between concentrations administered to squirrel monkeys and the doses appearing in Table 3 does not make sense.

183. p. 25, 5^th paragraph: No concentrations are predicted in Table 3. You might, however, calculate the Bq/L body water from the exposure to HTO to include in the comparison in Table 3.

ATSDR Response: This sentence has been re-phrased to read "However, these tritium concentrations are 25 to 1550 times higher than the maximum body water concentrations predicted by the doses in Table 3."

184. p. 27, first bullet, last line: What is a birth subject?

ATSDR Response: This is defined in the second sentence of the paragraph- "individuals born in Alameda County during the 1960 to 1990 period."

185. p. 27, bullets in general: There needs to be a break of some sort to make it clear to the reader that the first "sentence" is subject only so the reader need not look for a verb.

ATSDR Response: The bullets are used to denote a list of items.

186. p. 27 last bullet: In the follow up study for the years 1969 to 1980, were higher rates of melanoma found in addition to the cancers mentioned in the first sentence? Should sentence #2 start with "A" instead of "The"? Second sentence from end - "The later study" - Is this the Austin and Reynolds 1984 study instead of the 1980 one (Austin et al. 1981)?

ATSDR Response: The follow-up study was a more detailed evaluation of the cases identified in the 1980 study. It did not specifically evaluate rates of melanoma incidence. "Later" has been replaced with "Austin and Reynolds (1984)".

187. p. 28, top paragraph, last sentence repeats what was already said on p. 27.

ATSDR Response: The last two sentences of that paragraph have been deleted.

188. p. 29, 1^st paragraph, last sentence: How about "highly unlikely" because the 95% percentile is being used on top of a multitude of conservative assumptions?

ATSDR Response: "Unlikely" indicates an appropriate level of certainty regarding the public health determination of this evaluation.

189. p. 29 last paragraph: How are the potential doses from long-term releases included in the doses shown in Table 3?

ATSDR Response: The total tritium dose (column 6) is the sum of all of the dose components. The resulting total tritium dose, comprised of short term and annual doses, represents an annual dose for the year of the release (1970). These annual doses are compared with the ATSDR MRL, which is also an based on an annual dose.

190. p. 34-37: Appendix 1. In the text, the figures are ordered A4, A1, A2, no A3, A4 and A5.

ATSDR Response: The tables are re-ordered in the revised document.

191. p. 34, paragraph 4: As shown, the procedure for calculating the deposited HT is not correct either dimensionally or mathematically. Shouldn't it be that c/Q (s/m³) is multiplied by Q (167 Ci/s) times seconds of duration of release to get the values in Table A-2 (Ci-sec/m³), which are then multiplied by the deposition velocity (m/s)? The calculation is not presented clearly anywhere in the report.

ATSDR Response: The cumulative air concentrations in Tables A-2 and A-3 have been revised as a result of modifications in the assumed atmospheric stability classification and resulting dispersion calculations. The previously used version of RASCAL was a beta test version with debug features. That version produced dispersed air concentrations (both instantaneous and cumulative) in units of Ci/m³/Q. Multiplication by discharge Q produced concentrations in units of Ci/m³, which should be the stated units of Tables A-2 and A-3 (as those values were cumulative over a specified period, the time component was implied).

The current version of RASCAL does not include an option for directly outputting the calculated air concentrations. The output consists of cumulative surface concentrations in units of Ci/m². These cumulative concentrations are divided by the RASCAL deposition velocity (m/sec) to obtain cumulative air concentrations in units of Ci-sec/m³. The revised text and appendix describe this procedure.

192. p. 35: Why are the directions not symmetric around the centerline?

ATSDR Response: The plume distributions and concentrations are symmetric around the centerline. Output from the model may not produce symmetrical tables. The revised tables now indicate that the centerline of the 1965 plume is 60º and 20º for the 1970 release.

193. Table A-1: Why is there no column for 50? The information that distances < 0.7 miles are "on-site" is missing. Tables A-2 and A-3: Explanations of how to calculate deposition from "cumulative" are insufficient for a numerate reader who has little experience in dispersion modeling. Table A-3: Data need correcting, and the date is 01/20/65.

ATSDR Response: These tables have been revised, re-ordered, and re-labeled as suggested or indicated above.

194. p. 36: Table A-4: Data are missing at the bottom of the page. References to Table A-4 in the text use the term "summary information" which sounds like it will contain results, but it actually contains model inputs. In other words, don't replicate RASCAL jargon.

ATSDR Response: The RASCAL output labeled as "Case Summary" has been re-titled as "Input Summary". These input summaries have been revised due to newly available weather and release data.

195. p. 38: It is not made clear why the ISCST3 model can be used to model an emission of HT from a ground source (and not a stack). Further, the discussion of how the maximum values in Table A-6 are used is not very clear. For example, are these the maximum values that produce the results in Figure 4 on p. 14, and are how are these maximum values related to the Monte Carlo calculations for the 12-d exposure distribution (see p. 42)?

ATSDR Response: The ISC model requires hourly weather data in order model a specific release event. While we could use the same worst-case weather data as a proxy for the hourly 1970 data, this would ignore the data that are available and sufficient to run the RASCAL model. Conversely, the RASCAL model cannot model dispersion from an area source, such as results from the deposited HT plume. Figure 4 has been replaced and average and 95^th percentile initial (day one) HTO air concentrations have been added to Table 1. Table A-6 shows the similarity of the 5 weather years and that the concentrations from the worst-case year are used as a proxy for the 1970 data. This appendix has been re-written to clarify the uses and limitations of the air dispersion model.

196. p. 38, 4^th paragraph, line 6: There are no doses in Table A-6, only concentrations.

ATSDR Response: The caption refers to dose calculations "subsequent" to or derived from the listed concentrations.

197. p. 40: It is understood from statements on p. 11 (paragraph 1) and p. 22 (last paragraph), that ranges and most likely values for tritium dosimetry parameters were recommended by the expert panel review of tritium issues (citation = ATSDR 2002). This being understood, the equation identified on p. 40 should be attributed to both Cember (1988) and by reference to the ATSDR 2002 report too. Further this equation is paramount to the Monte Carlo simulations that are performed for deriving dose, but it identifies a tritium activity (Bq), which presumably represents body burden (e.g., see Cember, 1988, p. 359), as a concentration, and does not explain how that tritium activity is derived. Moreover, there is no definition for the [ddref or Ddref] term or the weighting factor. It is recommended that a better discussion of this equation be provided, and this equation be included in the main body of the text to show how the results presented in Figure 5, Table 2 and Table 3 might be computed.

ATSDR Response: The dose equation is from Cember (1988). Parameter values comprising that equation (as indicated by the bulleted highlights) are from the Expert Panel Report or other sources. References to the Expert Panel Report have been added to the bulleted descriptions for the radiation weight factor and the DDREF. See the response to comment 51 (page 54) for an explanation of the use of the tritium activity and body burden parameters. Discussion of the equation and associated parameters have been revised and expanded in Appendix 4.

198. p. 42-49 (Appendix 3): It would certainly be helpful if the figures presented as Forecasts and Assumptions appearing on these tables were labeled better, reflected the equations or other information used to create them, and referred to discussions, data tables, or figures in the text. What an outlier is should be explained.

ATSDR Response: This appendix has been re-written to clarify the evaluation process and assumptions and the forecast and assumption labels and titles have been checked for consistency with text descriptions as suggested.

199. p. 42: 12 day inhalation dose: The figure doesn't seem to match with the text. The display range goes to 200, not 300. The 95% probability shown is 0 to about 130, not 150.

ATSDR Response: The display ranges in the charts can be adjusted after the calculations are completed in order to better display the range of probabilities. Apparently, the labels are not re-generated with the chart.

200. p. 43: There's only a factor of 1.96 between initial soil concentration per m² and cumulative deposited HT, so it's not clear what period "initial" refers to.

ATSDR Response: The initial concentration is for day one and represents the maximum HTO soil concentration and subsequently, the maximum HTO emission rate from the soil.

201. p. 44: cumulative HT concentration (air): How do these results compare with those of RASCAL in Table A1? Are they at 0.7 mile?

ATSDR Response: The Monte Carlo simulations use cumulative HT concentrations (air) for each distance or location for which doses were calculated. Table A-1 contains instantaneous air concentrations and are not directly comparable. The mean of the cumulative HT concentration (air), which is 6.02 is equal, within rounding error, to the 5.95 value of Table A-2 at 0.7 miles (20º centerline). Revised assumption labels include the distance that they represent.

202. p. 45, Crystal Ball report: Cells D18 and K14 mean nothing to the reader. Their effect is 0, but as long as they were looked at, the reader should know what they are.

ATSDR Response: Cell D18 represents an inhalation dose from a light work breathing rate and K14 represents the strenuous work breathing rate. The Crystal Ball report and sensitivity chart have been updated and all of the significant assumptions and forecasts labeled.

203. p. 46: What is cumulative HT air concentration/Q? Cumulative air concentration should be Ci-sec/m³, and Q should be Ci/sec, so the units of this should be sec²/m³?

ATSDR Response: This is the original output unit of the RASCAL model. These values were multiplied by Q (Ci/sec) to obtain cumulative air concentrations in units of Ci-sec/m³ as suggested. As noted above, the revised RASCAL output is in terms of cumulative surface concentration (Ci/m²), which is divided by the RASCAL default deposition velocity (m/sec) to obtain cumulative air concentrations in the suggested units (Ci-sec/m³).

204. p. 46, HTO air conc: The mean of the distribution shown (2.1E-06 Ci/m³ = 2.1E-06 Ci m^-3 [not Ci/m^-3]) is presumably the maximum air concentration at 1 mile for the highest years (Table A-6). Is it correct to use a maximum value as the mean of a distribution?

ATSDR Response: The HTO concentrations values, as listed in Table A-6 are single values, or deterministic estimates of concentrations for different locations and weather years. We used concentrations from the weather years producing the highest concentrations because specific data for the 1965 and 1970 weather years are not available. It is reasonable to assume that deterministic results from the ISC dispersion calculations are estimated concentrations and that those estimates are as likely to underestimate the true value as they are to overestimate the true value. However, the revised calculations use the probability distribution of estimated HTO emission rates to calculate mean and 95^th percentile HTO emission rates which are used to estimate mean and 95^th percentile HTO air concentrations as described in the text and appendices.

205. p. 46: It would help to see a complete example of one Monte-Carlo iteration. That is, a list containing a single value for every input (although 5000 numbers is unrealistic, so some representative number of inputs), and a complete set of equations from which a single Monte-Carlo dose output can be calculated. This would provide a tool that readers could use to make sure they fully understand what was done.

ATSDR Response: The appendix describing the Monte Carlo calculations includes the specific equation and values of all of the assumptions used in estimating the doses.

206. Assumptions in general: Some additional discussion of the choice of distribution shape (normal, uniform, triangular, etc) would be helpful. If the real-world factors that contribute to deposition rate interact multiplicatively, then the lognormal would be a better choice. Empirically, one would expect asymmetric distributions for the milk concentration, and perhaps vegetation concentrations.

ATSDR Response: Discussions of the distributions of parameters (assumptions) are included in the appropriate sections and appendices. As previously indicated, these discussions have been expanded and the labels revised to clarify the calculation process.

The initial distribution of the tritium plume is evaluated via a Gaussian dispersion model. Concentrations from this model and consequently in soil and vegetation should be normally distributed. Nevertheless, the resulting tritium doses and many of the forecast distributions appear to be lognormally distributed which seems to reflect the real-world interactions.

207. Page iii states tritium effective half life varies from 10 to 40 days.

Page 40 states tritium effective half life ranges from 1 to 40 days.

ATSDR Response: The text on page iii has been changed to 1 to 40 days.

208. Throughout the document dose quantities from ICRP 60 methodology is discussed. Page 23 discusses doses using ICRP 26 methodology (i.e.. dose equivalent). Page 21 states that 40 CFR 61 limits doses to "an effective dose of 10 mrem/yr." It should state "an effective dose equivalent of 10 mrem/yr."

ATSDR Response: In ICRP 26 the term used is "dose equivalent." When the ICRP issued the new recommendations in ICRP 60, this term was replaced with the term "equivalent dose." Although the terminology changed, the method of calculation is essentially identical. The correct terminology, as you stated, in 40 CFR 61 is effective dose equivalent. We will correct the text as necessary.

Reviewer 5 Comments

209. Process problems. The health assessment process has been marked by a lack of responsiveness to community concerns, a series of contradictory documents, and very limited attention to establishing a record of what happened in the accidents and to informing the public in a detailed and understandable way about what happened. ATSDR has lost its opportunity to serve as an honest broker on these issues, and thus departed from its defined public health mission.

ATSDR Response: Upon initiating public health activities at the LLNL site, ATSDR established a site team composed of community members, local and state agency representatives, and other stakeholders in order to understand and address community public health concerns related to the LLNL site. The CDHS, in cooperative agreement with ATSDR, in a recently published a health consultation describing the Livermore communities public health concerns (CDHS 2003) specifically lists the prioritized concerns as defined by the site team. This public health assessment, along with numerous other ATSDR and CDHS documents specifically respond to those concerns raised by the Livermore community.

This public health assessment includes descriptions of both of the accidental releases and the respective environmental conditions during those releases based on all of the available information. The specific process by which we have estimated doses to the public from those releases has changed due to comments from the community as described in section two. Because the results of the dose estimation techniques are based on different assumptions and modeling approaches, they necessary produce different dose estimates. However, it is much more significant to note that for each of the modeling approaches, the resulting doses are below levels of public health concern and are consequently, not contradictory.

ATSDR has responded specifically and appropriately to all of the public health concerns related to the LLNL site. Dissatisfaction to our public health findings by a few activist groups is not necessarily reflective of the entire Livermore community. Based on numerous community presentations and meetings we are confident that the Livermore community accepts and supports our public health responses and determinations related to the LLNL site.

210. Treatment of uncertainties. These calculations involve a large degree of uncertainty, in part due to the unfortunate lack of information about the conditions around the accidents. ATSDR has made no evaluation of the reliability of the information derived from LLNL records on the release. Other key factors for which uncertainty was underestimated include meteorological conditions and the rate of tritium deposition. The treatment of uncertainty in the retention time of tritium in the human body is incoherent. There has been no attempt to explain in ordinary language the reasons for and the implications of the uncertainties in the modeling effort.

ATSDR Response: We acknowledge the uncertainty underlying the calculation of these dose estimates, which is why we have undertaken a probabilistic approach to the dose assessment. As stated in the introductory Note of Explanation, this document represents "the Agency's best effort, based on currently available information". Relative to the total magnitude of the tritium releases, existing LLNL information is the only available quantitative data such that there is no way to independently evaluate the reliability of that information. With respect to information about the conditions around the accidents, the revised calculations use newly available data on weather conditions and release parameters that should greatly reduce overall uncertainty of the resulting doses. Descriptions of the parameter assumptions underlying the dose calculations have been expanded in the appropriate sections and appendices (also see the response to comment 224 for an explanation of the uncertainty surrounding the biological retention of tritium).

Overall, we have attempted to balance known variability of environmental and dosimetric parameters with health protective assumptions about exposure and dose such that the estimated doses do not underestimate actual doses. As human measurements of tritium exposure at the time of the releases were not detectable, we are confident that the calculated doses are health protective estimates of the actual doses.

211. Calculation and presentation of dose estimates. The health assessment makes mistakes in presenting its results. In particular, the models predict higher rather than lower doses for the 1965 accident contrary to assertions in the text. A significant factor in calculating dose, the dose and dose-rate effect factor, was misused. We present revised dose estimates that correct these errors; our estimates are 3-4 times higher than those presented in the health assessment. Population dose estimates should have been made and we make a rough attempt at doing so here.

ATSDR Response: The estimated tritium concentrations and dose estimates for the 1965 release were appropriately estimated as less than those from the 1970 release. Although we have re-calculated both the 1965 and 1970 HT and HTO air concentrations on the basis of newly available data, the estimated concentrations from the 1970 release are still higher than those for the 1965 release. Consequently, we are continuing to use the 1970 dispersion calculations as the basis for the dose calculations.

The above referenced dose estimates (3-4 times larger than the ATSDR PHA estimates) are erroneously based on an exposure location at 0.5 miles from the tritium facility. This error (which is based on a typographical mistake in the public release PHA document) leads to the use of a multiplying factor of 2.83 (2 to 4 range). This multiplying factor accounts for virtually all of the (3-4 times) difference in the dose estimates. In light of the use of the newly available weather data, this use of a meteorological uncertainty factor (and/or a 0.5 mile exposure location) is not appropriate.

The dose and dose-rate effect factor, more precisely, the dose and dose-rate effectiveness factor (DDREF) is used to adjust the radiation dose to account for reduced effectiveness for radiation to induce cancer at low doses and dose rates, variabilities, and non-linearity of dose. The ICRP (1991) concluded that based on radiation biology results, linear dose is inconsistent with the data. The ICRP further evaluated the response and determined that a DDREF of 2 was appropriate to adjust for curvilinear effects associated with dose-effect relationships. In the correct application of the DDREF, the factor is an inverse factor; that is, it is used to reduce the dose. For an excellent discussion of the use of the DDREF, please see ICRP Publication 60 (page 69), the Radiation Effects Research Foundation web site http://www.rerf.or.jp or the National Radiological Protection Board of the United Kingdom http://www.nrpb.org for additional information.

The use of population estimates for the purposes of dose assessment has limitations as discussed in NCRP Report 121. The population estimates around the laboratory during the releases was based on a review of US Census information and physical evaluation using US Geological Survey maps showing structures in the direction of the plume. Our estimates indicate that the population was no more than 60 individuals and this number is too low to scientifically evaluate the dose to a population.

212. Discussion of health risks. In its treatment of risks from radiation exposure the authors of the health assessment contradict standard practice as described in the National Academy of Sciences BEIR V report (NRC 1995), in international commissions (ICRP 1991, UNSCEAR 2000), and the ATSDR Toxicological Profile for Ionizing Radiation (ATSDR 1999). The authors give no indication that their assumption of a threshold for radiation induced cancer is at variance with standard risk assessment practice or that there has been a very substantial scientific and policy debate on the issue. In contrast, using standard methods we find that within the range of uncertainty there was potential for cancer mortality risks that are considered 'significant' in common regulatory practice- that is, the risks using both ATSDR's and our estimates for a maximally exposed individual are in the vicinity of 1 in 10,000; in some uncertainty ranges, the risks exceed 1 in 1,000.

ATSDR Response: ATSDR, as outlined in its Cancer Framework Policy published in 1993 clearly states that ATSDR does not perform risk assessments. Therefore, the public health assessment was not designed to evaluate the risks associated with releases of tritium. As the Cancer Framework Policy states "algorithmically derived numerical risk estimates tend to be conveyed in an artificially precise manner and sometimes used inappropriately in decision-making."

The currently accepted risk estimate for radiation induced cancer is 0.0005 per rem per year. However, the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) suggests that uncertainties in cancer risk estimates may be about twofold higher or lower for acute doses where cancer risk can be directly assessed and a further factor of two (higher or lower) for the projection of these risks to very low doses and low dose rates. Therefore the uncertainty can range from about 0.000125 to 0.002 per rem per year.

The maximum dose estimates in the public health assessment are about 40 mrem to an adult. Using the 0.0005 per rem estimate, this results in an annual estimated risk of 2 in 100,000, well within the EPA risk range of 1 in 10,000 to 1 in 1,000,000. For a child who may have received a dose of 140 mrem, the associated risk is about 7 in 100,000. Because you did not supply your evaluation of uncertainty issues, we cannot respond further to your statements about potential risks exceeding 1 in 1,000.

As discussed in the ATSDR Toxicological Profile for Ionizing Radiation, the scientific literature is very clear that a radiogenic threshold exists for the induction of cancer. This is clearly seen in the atomic bomb survivors, radium dial painters, uranium miners, and medical patients. The linear non-threshold (LNT) THEORY is appropriate for use in setting regulatory standards with necessary safety factors. However, the LNT theory is not a valid scientific model for evaluating health effects for environmental exposures. This is also discussed in the ATSDR Cancer Framework Policy.

213. Irresponsible conclusions. The assessment and the consultations use the term "below levels of public health concern" in a number of places, including in its conclusions about potential risks. There are serious problems with this usage. The term is nowhere defined, nor is there any indication of what the authors would consider to be a level of public health concern. Risks calculated using standard practice from the radiation doses presented in the assessment are at levels that are generally taken to be significant by the agencies supervising Superfund clean ups. Most disturbingly, the inferences drawn by ATSDR directly subvert the ALARA principle (as low as reasonably achievable), a cornerstone of the social compact for managing radiological hazards. The impression left by the ATSDR documents is indifference to releases of 300,000 Ci of tritium (in the form of hydrogen, or 10,000Ci in the form of water vapor) in a highly populated area.

ATSDR Response: The phrase "no public health concern" has been replaced with the phrase "below levels of public health concern". Further, "levels of public health concern" are clearly defined by addition of a subsection on "Tritium Doses of Public Health Concern" to the Public Health Implications Section. This subsection clearly identifies the health comparison values and standards that ATSR uses in making its health determinations.

ALARA (As Low As Reasonably Achievable) as defined by the US Nuclear Regulatory Commission "means making every reasonable effort to maintain exposures to ionizing radiation as far below the dose limits as practical, consistent with the purpose for which the licensed activity is undertaken, taking into account the state of technology, the economics of improvements in relation to state of technology, the economics of improvements in relation to benefits to the public health and safety, and other societal and socioeconomic considerations, and in relation to utilization of nuclear energy and licensed materials in the public interest" (see 10 CFR 20.1003). The ALARA concept is not relevant for retrospective dose assessment such as performed by ATSDR in this public health assessment.

Our public health determination that the tritium releases posed "no apparent public health hazard" does not represent indifference to the public health issue. This finding is based on a quantitative evaluation that the resulting exposures did not occur at doses likely to cause any adverse health affects and were consequently "below levels of public health concern."

[The following paragraph references relate to Section 4, specifically the subsection on "Toxicology of Tritium]

214. Paragraphs one and two. We agree that animal studies of tritium effects and human and animal studies of the effects of exposure to gamma radiation and x-rays are the only sources (and good sources) of data bearing on the effects of tritium in humans. Of course, we also believe that where human data are available they should be the principle evidence considered.

ATSDR Response: We agree that human health studies are the best source of information for determining the health effects of tritium exposure. It is important to note that there have been many reported cases of human tritium exposure, but there have been no documented observations of adverse health effects related to those exposures. Consequently, the potential human health effects from tritium exposure must be extrapolated from other sources of information.

215. Paragraph three. The statement that "the lowest tritium dose capable of causing adverse health effects is 3 rad" is contrary to the no threshold approach and an irresponsible claim. Even if we consider this to be a dose below which no health effects have been observed we find that it is simply not true (see below). Furthermore, this value is based on mice exposed in utero when there are several studies available that address in utero exposures to low-dose, low-LET radiation in humans.

ATSDR Response: Specifically, we are not making that statement, we are referencing a scientific study that has observed a threshold for adverse health effects for specific cases. Further, many studies of nuclear workers, patients receiving nuclear medicine procedures, radium dial workers, and others have shown a dose response in which no adverse health effects were observed at doses less than 5 rads (or 5 rem). For a general review, please see the General Accounting Office (2000) review of radiation standards. The linear-no threshold approach is actually the linear no-threshold theory (LNT) which is used in setting radiation standards. What is not widely known is that the LNT is based not on human studies but on x-ray studies on male fruit flies. In studies with female flies, no response is seen below 80 rem (http://mailman.mcmaster.ca/mailman/private/cdn-nucl-l/0201.gz/msg00111.html). The LNT does not take into account repair mechanisms, target theory, direct and indirect action of radiation, or other mechanisms that have conclusively been shown to affect the cellular and organism response to ionizing radiation. As for the statements about mice and low LET radiation, studies initiated by the Russells led to the findings that the dose necessary to double the mutation rate in mice receiving protracted (Low Dose-Rate) radiation is on the order of 90 to100 rads. Therefore, our statement is based on a review of the scientific data associated with radiation response in the dose range observed around LLNL. With regard to in utero studies of humans, we have based the review on those studies reporting the lowest doses, whether human or animal.

216. Paragraphs four and five. In these paragraphs the authors describe a linear no-threshold dose-response for negative effects on intelligence after in utero exposures. As stated above, we consider this to be the standard model for cancer and can easily conceive of such a relationship existing for other health effects as well. Unfortunately the authors draw an erroneous conclusion. At doses being considered here, intelligence would only be changed by a fraction of a point on the scale used to evaluate the A-bomb survivors. The consultation states that since this would not be measurable there would be no effect. However, even effects that can't be measured can still be present. In this case, if the model being cited were correct, there would be a small effect on intelligence. At low doses such an effect would not be measurable at the individual level but if sufficient numbers of people were exposed, it could manifest itself as a change in the distribution of intelligence based on a similar scale (more students requiring remedial education or fewer 'honors' student). We should also mention that if you consider lost intelligence to be a health effect than this admission of a no-threshold dose-response contradicts the threshold mentioned in paragraph three.

ATSDR Response: Although it is conceivable that radiation could cause some effect that cannot be seen, that does not mean it is deleterious. There is sufficient weight-of-evidence in the literature to show a protective effect of low level radiation; that is, a hormetic effect. In fact, the US Court of Appeals ordered the US EPA not to use LNT in a chemical carcinogenesis case (Chlorine Chemistry Council v. EPA (D.C. Cir., March 31, 2000)). As a result of this court ruling, EPA is re-evaluating across the board, its use of LNT in risk assessments, recognizing that in many cases, LNT may not be applicable especially as it relates to ionizing radiation (for reference, see http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=55445). In fact in 40 CFR 197, EPA states that radiation effects, depending on the dose, can be either stochastic (effects occur by chance or no threshold) or non-stochastic (threshold).

The use of intelligence as a lowest observed adverse effect level for acute exposures is discussed in the ATSDR toxicological profile on ionizing radiation. In summary, the ATSDR MRL for acute exposures to ionizing radiation is 400 mrem. The atomic bomb survivor studies showed there was a difference in intelligence tests when the in utero acute doses exceeded 25 rem (25,000 mrem). The tritium doses from these accidents were also acute exposures thus we believe there would be no observable decrease in intelligence.

217. Paragraphs six and seven. These paragraphs deal with cancer following adult and in utero exposures, respectively. The threshold for leukemia after adult exposures is placed at 100 rad and the threshold for cancer after fetal exposures is placed at 10 rad. Lower doses of 20 rad (adult) and 1 rad (in utero) are mentioned but disregarded.

Again we would like to point out that the consensus about radiation carcinogenesis is that there is no threshold for cancer. That said, the field of epidemiology has provided numerous examples of cancers attributable to radiation doses well below 20 rad. These are plentiful enough to have been reviewed several times (for example Ron 1998, Schbauer-Berigan and Wenzl 2001). In one of the most comprehensive analyses of existing literature that has been undertaken Wilkinson and Dreyer (1991) pooled the data from seven nuclear worker cohorts in the US and the UK and detected an excess of leukemia in the group exposed to between 1 and 5 rem. The A-bomb survivorship have also shown excess cancers at low doses; Pierce and Preston (2000) point out that most of the survivors had doses less than 20 rem. They emphasize that " there is direct, statistically significant evidence of risk in the dose range of approximately 0-0.10 Sv (0- 10 rem)". Clearly a 100-rad threshold for cancer following adult exposures is unjustifiable.

Cancer following in utero exposures has also been extensively studied and the body of evidence reviewed. The pioneering work of Alice Stewart with prenatal x-rays (Stewart et al 1956) grew into the Oxford Survey of Childhood Cancers and its results were quickly reproduced in the US (MacMahon 1962). Much more evidence has piled up supporting sensitivity to radiation during fetal development and one of the recent reviews concludes that "radiation doses on the order of 10 mGy (1 rad) received by the fetus in utero produce a consequent increase in the risk of childhood cancer" (Doll and Wakeford 1997).

ATSDR Response: We have not disregarded the in utero doses, we have specifically indicated how much lower the LLNL estimated doses are relative to those in utero doses which have been found to cause adverse health effects. Based on the NRC recommended dose adjustment for fetal exposures (fetal doses = 1.7 times the dose the mother), fetal doses from the LLNL accidental releases would be less than 71 mrem (95^th percentile) for the maximally exposed individual fetus. This estimated maximum dose is more than 10 times lower than the quoted dose of 1 rad (1 rem).

There is not a consensus for LNT either for chemical or radiological carcinogenicity. The ATSDR Cancer Framework and the EPA Cancer Risk Assessment policies recommend use of a weight of evidence approach rather than rote application of the LNT theory for cancer evaluation. Both of these policy documents have been peer reviewed and may be considered to represent a consensus approach to cancer risk evaluation. We agree that epidemiological studies have indicated radiation induced cancer at the doses you discussed (in excess of 10 rads); however, the doses associated with the tritium releases are a fraction of a rad.

More recent studies by the American Association of Physicists in Medicine (AAPM) estimate the risk of adverse effects of radiation exposure on a unborn child at less than 1 in 1,000 for 1 rem of exposure. They also state that the natural risk of congenital defects is much greater than the effects associated with the additional risk from the x-rays. Although the work by Stewart was groundbreaking, we believe it has been mis-interpreted. According to AAPM, "...both the American College of Radiology and the American College of Obstetrics and Gynecology have adopted a policy that rarely, if ever, is termination of pregnancy advisable because of the radiation risk arising from diagnostic x-ray examinations."

218. Paragraph eight. This paragraph suggests that low doses of radiation might increase the lifespan of exposed individuals and that the doses received downwind of LLNL would not shorten lifespan. We acknowledge that there are proponents of this life-increasing hypothesis; this is a controversial suggestion, evidence is not coherent on this issue, and it is in any case a distraction.

ATSDR Response: Considering that the adverse health effects from low radiation doses are based on linear extrapolations of high dose effects, it is significant that positive health effects have been observed from low dose exposures. This paragraph specifically addresses a prior comment that decreased lifespan was of concern to community members.

219. Paragraph nine. This introduces the idea that women and children are more vulnerable to the effects of radiation exposures than the reference man. We agree that children, exposed to the same air concentrations as an adult, will receive a higher dose, and this probably true of fetal doses as well. This is explained in the dosimetry section of Appendix A. It may also be the case that a woman will receive a higher dose than a man based on a smaller average body weight (Richardson et al 2001a).

Higher doses are half of the problem. The other half of the problem is increased sensitivity to a given dose. Human exposures early in life (in utero or as a young child) and late in life (over age 50) have been associated with higher risks of cancer, although this is complicated by the fact that the pattern of risk seems to be different for different cancer sites (Richardson et al 2001b, Ritz et al 1999). There is also animal evidence for this effect (Walinder and Sjoden 1973, Sasaki et al 1978, Di Majo et al 1990, Benjamin et al 1991, Sasaki 1991, Van Den Heuvel et al 1995). Another possible sensitivity was described by Nomura (1984) looking at mice. This study observed that fetal x-ray exposures made mice hypersensitive to chemical-induced lung cancer later in life, an effect not observed with postnatal x-ray exposures. Despite the extensive literature on this topic, the increased sensitivity of young and old individuals to radiation was not explored in the consultation.

ATSDR Response: As we did not receive the complete references for those you cite, we cannot thoroughly comment on your statements. However, the Nomura study incorporated a dose of 36 rads or 100 times the typical background dose in the United States. The study showed that after irradiation, the induction of lung cancer following a chemical agent (urethane) was increased. Nomura also stated that the in utero irradiation was not tumorigenic. The Di Majo study used neutrons which are inherently more dangerous that tritium. This study stated that there were small effects seen with x-ray doses as low as 30 rads. Following irradiation with neutrons, the effects were seen at doses as low as 17 rads. This study also irradiated animals at doses well in excess of background and well in excess of the doses calculated around LLNL. These studies cannot be extrapolated down to the doses around LLNL.

We agree that there are radiation sensitivities with regard to age and sex as well as ethnicity. In fact, this topic has been discussed extensively in the literature as it is of paramount importance that the dose estimates from the atomic bomb survivors be correctly modified to other ethnicities. The ATSDR MRL, which we are using as a screening value for potential adverse health effects, specifically considers especially susceptible portions of the population.

220. Paragraphs 10 through 13. These paragraphs mention genetic effects in a muddled way. Genetic effects, also called heritable effects, arise from mutations in male or female reproductive cells and are expressed in the offspring of those exposed. Sperm count reduction and effects of in utero exposures are described although they are not genetic effects. This consultation does not in fact mention any of the evidence for genetic effects, much less discuss the magnitude of genetic risks, despite an enormous amount of available information (Chapter 2 of BEIR V is devoted to the topic). Animal data is of course the bulk of the evidence, including studies that show heritable cancers (Nomura 1982, Mohr et al 1999). There are also a few human epidemiological studies, including a controversial study of leukemia in the offspring of exposed nuclear workers (Gardner et al 1990) and a robust pair of experiments looking at germline mutations around Chernobyl and the Semipalatinsk nuclear test site (Dubrova et al 1996, 2002). Much of this evidence is difficult to translate into a risk estimate, but Straume (1993) methodically laid out the categories of genetic risk and estimates a comprehensive genetic risk based on the data in BEIR V. This number could have been easily used in the consultation to present a ballpark genetic risk estimate for the accidents.

ATSDR Response: Comment noted.

221. The final paragraph concludes that "the doses to all members of the Livermore community are not at levels of public health concern". We find this statement to be as illgrounded as the discussion that precedes it; furthermore, the doses and risks in question may have been relatively low, but to dismiss community concerns so trivially is nothing less than an insult.

ATSDR Response: The purpose of the public health assessment is to evaluate the public health hazard posed by the accidental tritium releases that occurred in 1965 and 1970. We have made health-protective estimates of the possible doses that could have been received by members of the Livermore community and determined that those doses were too low to produce any adverse health effects. We have fully addressed the communities concerns regarding these potential doses and conclude that those historic tritium releases did not result in a public health hazard and consequently, are below levels of public health concern.

222. The ATSDR assessment (ATSDR 2002) appears to have identified the most significant pathway for exposure from releases of tritium in the form of hydrogen; it is the inhalation of tritiated water vapor released after capture of the tritiated hydrogen by soil and vegetation. This is an important finding and merits emphasis. It contradicts the approach used in the earlier health consultation (ATSDR 2001a). The assessment dose estimates for nearby exposures are surprisingly large compared to those in the consultation, but they are plausible. The discussion of uncertainties in the dose estimates, however, is limited and misleading.

ATSDR Response: We also believe that HTO inhalation is a significant route of human exposure to the released HT. However, based on additional comments, we have included a dermal absorption component that is equal to the HTO inhalation dose. It is also important to point out that the inhalation doses were estimated using 1 hour and 12 hour (revised version) maximum averages that probably overestimate doses derived from a 24 hour or daily concentration (the maximum 24 hour averages are about of the maximum 12 hour averages). We agree that discussion of the probability distribution of the estimated doses is limited. Ultimately, we must ensure that the public health determinations are based on estimated doses that do not underestimate potential historic doses. We are confident that no one in the Livermore community received a tritium dose from either the 1965 or 1970 releases that was greater than those estimated in this public health assessment.

223. The unavailability of detailed information about both accidents is very unfortunate. No information has appeared about the 1965 accident so we do not even know what direction the wind was blowing, whether LLNL made any radiological measurements, or had any sort of emergency response. One article after the 1970 accident (Myers et al.) describes the meteorological conditions and a significant effort at monitoring as a follow up to the accident. No detailed work has been done to combine modeling with the monitoring report to reconstruct a better picture of what happened after the release.

ATSDR Response: LLNL recently released portions of the accident report relating to the 1965 release. Data from that report, as well as additional weather data, have been incorporated in the revised dispersion and dose calculations. As the environmental analyses conducted after the 1970 release have limited geo-locators and sample time indicators, we feel we have used that data to its fullest extent.

224. The assessment gives an inadequate treatment of key uncertainties in the modeling effort. The most important of these are: 1) a failure to consider the implications of uncertainties in the actual meteorology at the time of the accidents: a particular wind speed, direction, and stability assumption is used for the 1965 accident without considering what doses might be expected for alternative conditions; uncertainties in stability conditions could be significant for modeling the 1970 accident as well; 2) deposition velocities are more uncertain than the three-fold factor considered by ATSDR and they directly affect the availability of tritiated water vapor for inhalation; 3) the treatment of retention time in the dosimetry is incoherent: it is not the case that we don't know whether retention times are 1 day or 40 days; rather we know that there is a short and long component to retention times, that there is some variation between people in the relative importance of those components, and we have some uncertainty about times and relative importance within this picture; 4) similar incoherence appears in the treatment of other dose-affecting parameters like breathing rates, activity patterns, and body weight; there is variability in these, but not much uncertainty and the modeling should reflect this.

ATSDR Response: In response to newly available weather data we have revised the air dispersion calculations and the resulting dose estimates. We have run the dispersion calculations using several different stabilities for each release and estimated the doses using the stabilities leading to the highest potential air concentrations in areas of maximum exposure. As suggested, the revised weather data did significantly affect the dispersion pattern of the respective plumes. However, the cumulative effect of revisions to the dispersion calculations and HTO loss rates resulted in very minor changes in the estimated doses such that the overall public health findings have not changed.

We have also included additional references to HT deposition velocities. However, these additional references support the assumed distribution of velocities such that we have not revised this parameter. With regard to the biological retention of tritium, the ICRP model for HTO recommends a clearance rate of 97% after 10 days with the remaining 3% excreted after 40 days (as referenced in ATSDR 2002). For OBT, the respective rates are 50% after 10 days and 50% after 40 days. We have combined these rates into a single distribution that varies from 1 to 40 days, with a most likely value of 10 days, and an average value of 17.7 days. The resulting distribution of retention times is a health protective estimate of the biological half life of tritium in the human body.

225. The health assessment gives a description of "conservative" or "health protective" assumptions that clouds the interpretation of its findings, especially since there is no complete or coherent treatment of uncertainty and no assessment of the magnitude of the conservatisms.

ATSDR Response: If the estimated doses are both greater than the actual or historical tritium doses and less than doses likely to cause adverse health effects, then the magnitude of the conservatisms is irrelevant for the purpose of public health assessment. However, we have also presented the estimated doses in terms of average values and 95^th percentile values. This range is an indication of the uncertainty of the estimated doses.

226. The health assessment makes mistakes in presenting its results. In particular, the models predict higher rather than lower doses for the 1965 accident contrary to the assertions in the text and tables.

ATSDR Response: Both the public comment and final version of the PHA are correct in predicting that air concentrations and the resulting doses were higher for the 1970 release than for the 1965 release. There was a typographical error in the previous version indicating the exposure to the 1965 release occurred at a distance of 0.5 mile from the source. The nearest residential exposures for both releases was at locations more than 1 mile from the source.

227. The assessment only calculates doses to "maximally exposed individuals" at various distances close to the accident. We have made some very crude sample estimates of doses to populations at the regional and global spatial scale. These are 15-300 person Rem for within 50 miles, 70 -1500 person Rem within the U.S.and, very roughly, 3000 - 20,000 person Rem in the Northern Hemisphere (including the U.S.) in later months and years.

ATSDR Response: Adverse health effects from exposures to hazardous substances are based on exposure and intake of the substance to a real person. These exposures and intakes of hazardous substances can be estimated as an individual dose to the most sensitive portion of the exposed population. If the maximum estimated dose to the most sensitive individual is unlikely to result in any adverse health effects, then any lower dose is also below a level of public health concern.

228. The assessment provides its own interpretation of the literature on radiation-related health effects. This interpretation is inconsistent with standard practice, exemplified by the ATSDR toxicological profile on Ionizing Radiation (ATSDR1999) and the National Academy's BEIR V report (NRC 1995). The assessment discussion does not relate its approach to standard practice, nor does it consider other alternative interpretations of the literature.

ATSDR Response: Please see the response to comment 212.

229. The assessment and the consultations use the term "below levels of public health concern" in a number of places, including in its conclusions about potential risks. There are three serious problems with this usage:

1) The term is nowhere defined, nor is there any indication of what would be a level of public health concern.

2) Risks calculated using standard practice from the radiation doses presented in the assessment are at levels that are generally taken to be significant by the agencies supervising Superfund clean ups.

3) Most disturbingly, the inferences drawn by ATSDR directly subvert the ALARA principle (as low as reasonably achievable), a cornerstone of the social compact for managing radiological hazards.

ATSDR Response: See above responses to comments 112, 212, and 213, respectively.

230. To begin with, we find no clear statement of what questions are being answered within the assessment, why these answers address stakeholders' need for information, or why the calculations are done the way they were.

ATSDR Response: Paragraph 3 of the introduction includes the following sentence. "During the course of the LLNL public health assessment process, potential off-site exposure to tritium released by LLNL was identified as a specific community concern (CDHS in review)." In order to clarify this statement of purpose, the following sentence has been added. "Consequently, this public health assessment will specifically evaluate the potential for adverse health effects in the Livermore community from the accidental tritium releases." Explanations of the calculation processes and underlying assumptions that are provided in section 3 and appendices 2-4 have been revised in response to specific comments.

231. As we have noted above, continuity among the documents is lacking. New documents that represent ongoing agency analysis should explicitly address their relationship with previous work. When new approaches are used or there are new findings, the reasons for the changes should be explained and placed in a context of ongoing work. In the absence of such explanations the agency creates confusion and undermines trust.

ATSDR Response: Section 2 of the PHA is a description of the comments that ATSDR has received on both the earlier health consultation and the public release version of the PHA. As stated in that discussion, there were several comments that elicited extensive revision of each document. The initial approach used in the health consultation was an attempt to update the Myers et al. (1971) study using the RASCAL dispersion model. As stated in the comments to that consultation (section 2 of the PHA), that approach did not adequately address the environmental fate and transport of the dispersed HT plume. Consequently, an expanded PHA was developed that explicitly addressed all of the comments. However, because of the inherent differences in modeling approach used in the documents, there is no basis for comparing results. Comments on the public release version of the PHA have also elicited a number of revisions, however, the basic modeling approach and estimated doses are both comparable and similar.

232. There is very little evidence within the assessment of the "give and take" in problem definition and interpretation of results that one would expect of ananalytic and deliberative process involving stakeholders. Recommendations should be developed through a collaborative process involving stakeholders.

ATSDR Response: The evolution of this public health assessment, as described in Section 2 and indicated above, belies the statement that "There is very little evidence within the assessment of the 'give and take'" Both the dose estimation process and the resulting dose calculations have been extensively improved as a result of community input and comment. Public health recommendations are only developed as warranted by conditions or exposures of public health concern. As the estimated doses are below levels expected to cause adverse or detrimental health effects, no recommendations are warranted.

233. The stakeholders are entitled to a more complete story of the accidents with some outside review of the information provided by LLNL. ATSDR could facilitate this.

ATSDR Response: ATSDR has submitted this health assessment to outside peer review. Peer reviewer comments are included with appropriate responses.

234. A health study that would directly relate health effects specifically to the accident releases is not feasible.

ATSDR Response: We agree that tritium doses representing a fraction of normal background exposures will not result in observable adverse health effects.

235. There are lessons that could and perhaps should be drawn from the accidents regarding emergency planning, environmental monitoring, and, conceivably, health monitoring.

ATSDR Response: Improvements in emergency planning and environmental and health monitoring programs over those from the 1960's and 70's are well documented in annual reports produced by LLNL.

236. Dose Parameters:

This section briefly describe the parameters used to estimate doses received. The equation as shown on page 40 of the ATSDR Assessment is:

Dose (Sv) = Tritium concentration (Bq) x Energy of Tritium beta decay x J/MeV x sec/day x DDREF x wt factor/body mass)/lambda;

Where DDREF is a dose and dose-rate effect factor, the wt factor serves to account for relative biological effectiveness (RBE) and lambda is equal to the ln(2) divided by the biological half-life of tritium.

The equation assumes a certain amount of radioactivity and calculates the energy absorbed with each radioactive disintegration. It also calculates how many disintegrations will happen before all of the tritium has left the body. Dose is calculated by dividing the energy delivered by all of these doses by the body weight, leaving a certain amount of energy per kilogram of body tissue. This dose is adjusted by certain factors, in this case DREF and RBE.

We looked at an alternative tritium dose calculation model presented by D. M. Hamby, one of ATSDR's expert panel. Hamby's equation was written as:

Dose (Sv) = (f₁*QF*Ã*E*M_s
-1*T_B)/ln(2);

where DCF is the dose conversion factor, f₁ is fractional absorption, QF is quality factor (or RBE), Ã is the integrated activity intake, E is average beta energy per disintegration, M_s is soft tissue mass, and T_B is biological half-life. The major difference in the two equations is the choice of adjustment factors; Hamby adjusts the dose upward with a tritium QF (RBE) of between 1 and 3.5. The ATSDR equation uses a factor (the wt factor, used in the same way as RBE) of 1.3. The ATSDR equation also adjusts the estimate downward with a DDREF of 0.4. We take issue with this use of the DDREF. The expert panel report used the DDREF in a risk equation, and we agree that this is where such a factor belongs.

ATSDR Response: The radiation weight factor distribution used in our simulations is based on the stated range and most likely values recommended by the expert panel. The values (in a triangular distribution) ranged from 1 to 3, had a most likely value of 1.3, and an average value of ~1.8. The resulting radiation weight factor is more conservative than the ICRP recommended quality factor of 1. The dose and dose rate effectiveness factor (DDREF) is used to extrapolate from high doses (derived from A-bomb studies) to low dose and dose rates. Typically, the DDREF is applied when the doses are less than 20 rads as recommended by the UNSCEAR 1993 report.

Reviewer 6 Comments

237. ATSDR produced a draft Public Health Assessment (May 24, 2002) on the impact of tritium released from LLNL in 1965 and 1970. The report includes discussions of environmental pathways, community health concerns, community exposures, and public health implications. ATSDR's estimate of the public health impact of tritium released from LLNL is flawed by a biased review of the literature on biological effects of low-level radiation, misinterpretation of epidemiologic principles, omission of information from studies of humans exposed to tritium, and a lack of consideration of important determinants of susceptibility to radiation exposures.

ATSDR Response: Please see previous or following responses to specific comments.

238. Although the ATSDR report states that ionizing radiation from tritium can have health impacts including cancer and genetic effects, it interprets animal and human studies of gamma radiation and x-rays as showing there is a threshold dose for health effects of tritium. This interpretation is at odds with biological theory and recent evaluations of evidence from in vitro, animal and human studies (1-3). The energy in a single beta particle produced by radioactive decay of tritium is sufficient to cause ionization in a cellular environment and thereby directly or indirectly cause damage to genetic material. This means that, on theoretical grounds, there is no threshold for biological damage. The capacity of low level radiation doses from tritium to produce health effects is dependent on the population dose and the susceptibility of the individuals exposed, which may be related to their developmental stage, the effectiveness of genetic repair mechanisms, other carcinogenic exposures, reproductive behavior, and other factors. The National Academy of Science's BEIR V committee concluded that it is reasonable to assume there is no threshold dose below which health effects of ionizing radiation do not occur (2).

ATSDR Response: Please see above responses to comments 212, 216, and 217. Furthermore, the EPA has recently promoted the use of a weight of evidence approach to cancer risk assessment relative to the use of the linear non-threshold approach in dealing with cancer risk assessments (EPA 2003).

239. ATSDR relies on studies of laboratory animals to provide assurance that tritium exposures from LLNL releases did not produce health effects. ATSDR fails to note that laboratory animals have life spans that are shorter than the latency periods for many human cancers. There can be large inter- and intra-species variation in susceptibility to carcinogens. Although animal studies are important, their relevance to effects of radiation on heterogeneous human populations in complex environments that include exposures to other initiating and promoting agents should not be over-stated.

ATSDR Response: Please see the response to comments 214 and 215.

240. It is increasingly clear that there are fundamental flaws in studies of the survivors of the atomic attacks on Hiroshima and Nagasaki (4-11). High mortality in the aftermath of the bombings, selective survival of persons with low sensitivity to radiation effects, and uncertainties in estimating doses from the bombs, including a lack of data on doses from radioactive fallout, raise serious questions about the validity of risk estimates from this population. The ATSDR report relies on risk estimates from studies of the A-bomb survivors without acknowledging that there are several reasons why these studies should be expected to underestimate radiation health effects. Furthermore, the report misinterprets analyses of risk in arbitrary dose groups, fails to point out that standard dose response analyses which avoid arbitrary dose categorization suggest no threshold for radiation induced cancer (2), and fails to cite recent reports which support a linear no-threshold model for radiation health effects (12).

ATSDR Response: Please see the responses to comments 212, 216, and 217.

241. Beginning during the Manhattan Project to develop atomic weapons, workers exposed to chronic, low level ionizing radiation have been monitored using dosimeters incorporated into security badges. Dosimeters have been used primarily to identify excessive exposures, however, beginning in the 1960s, badge readings were assembled for studies of radiation risks. Some early studies emphasized low death rates of nuclear workers compared to national statistics (13), however, low mortality among workers can be misleading because death rates are typically low among employees of large corporations, which conduct health screening and offer benefits of employment, compared to the general public, which includes people who are too sick to work. Other research has examined cancer death rates of workers according to the levels of radiation recorded on their badges. A growing number of studies that carefully evaluate radiation measurements and other worker characteristics have shown that cancer rates rise with increasing radiation exposure, even at dose levels permissible under current stndards (10, 14-29), adding to other literature that suggests there is no threshold for carcinogenic effects of ionizing radiation.

ATSDR Response: As with other organizations, ATSDR is aware of these limitations on the dosimetry system (DS86) used in Japan. We hope that the new dosimetry system being prepared will correct some of these shortcomings. However, no studies have ever documented any adverse health effects, including cancer, from radiation doses near or approaching background levels (~360 mrem/year). In fact, no peer-reviewed studies have shown a clear relationship of radiation exposure to adverse health effects at doses up to 1 rem. The United Nations has reported areas of the world where the background is in excess of 1 rem per year with no apparent adverse health effects. The estimated doses to the Livermore community are a fraction of background levels.

242. Although most of the whole body radiation exposures of nuclear workers comes from external penetrating radiation, some workers have been exposed to tritium, which distributes throughout the body. Where nuclear facilities have monitored tritium exposures by urinalysis, tritium doses can be combined with external radiation doses to yield a total whole body dose. Several epidemiologic studies have demonstrated relationships between these combined whole body doses and mortality from leukemia, multiple myeloma, and other cancers (29, 30).

ATSDR Response: The term total effective dose equivalent (TEDE) is used to combine both internal and external radiation exposures and the resulting radiation dose. With those studies, it would be difficult to control for the exposure to external radiation thus determining the specific effects due to the tritium exposure. In recent studies of people wearing tritium powered watches, no adverse health effects were observed with tritium in urine concentrations ranging from over 5,000 to 30,000 picocuries per liter. Following the 1970 LLNL release, workers and community residents did not have detectable concentrations of tritium in urine samples (detection limits were 5,000 pCi/L).

243. The availability of individual radiation measurements for thousands of workers whose cancer mortality has been monitored over decades is a unique epidemiologic resource that does not exist for any other occupational exposure (10). Although workers are generally healthy, they are more similar to the average members of the public than are subjects of some other types of radiation epidemiology studies, such as patients under treatment or survivors of atomic attack. These studies are therefore especially important to consider in evaluating public concern about radiation releases. ATSDR does not discuss occupational radiation studies in its draft report.

ATSDR Response: We have made no specific references to worker studies because this health assessment does not specifically address worker-related exposures. However, the results of worker studies have been extensively reviewed in the ATSDR Toxicological Profile on Ionizing Radiation (1999). Even though worker exposures are typically much greater than the off-site community exposures evaluated in this PHA and may not be applicable to children, the information relating to worker studies has been considered in the development of the ATSDR MRL for ionizing radiation.

244. In the early 1950s, Alice M. Stewart began the Oxford Survey of Childhood Cancers. British children who died of cancer were compared to healthy controls in an attempt to identify exposures that might be causes of childhood cancer. In 1956, the first of many reports from the OSCC, which became the largest study of childhood cancer ever to be conducted, showed that the only factor that differed systematically between cancer cases and controls was a history of maternal pelvic x-ray during pregnancy (31). Because of the widespread assumption in the medical community that low level radiation posed no risk, in part due to the absence of an effect in the study of in utero exposed A-bomb survivors, the initial OSCC findings were treated with great skepticism. However, the medical x-ray link to childhood cancer held up as additional cases and controls were added to the OSCC (32, 33), and subsequent studies in other countries confirmed the British findings (34). Exposures of pregnant women to diagnostic x-rays have been greatly curtailed even though the doses from diagnostic x-rays are well below the levels that ATSDR cites as posing no risk.

ATSDR Response: Please see the response to comment 217. Additionally, the historic practice of maternal pelvic x-rays involved much higher doses than current diagnostic techniques.

245. Although the issue of in utero radiation exposure is raised in the ATSDR report, there is misinterpretation of the literature, perhaps due to reliance on a health physics text rather than a review of the literature on the carcinogenic effects of obstetric x-rays. Extensive evidence now shows that childhood cancers in general (not just leukemia, as cited in the ATSDR report) are related to fetal exposure to diagnostic x-rays. Mean whole body fetal dose estimates in these studies are approximately 0.6 rad, well below the 10-50 rad figure cited by ATSDR. (35, 36)

ATSDR Response: Estimated mean whole body fetal doses from the 1970 accidental release (to the maximally exposed individual) are less than 20 mrem. This dose is more than 30 times lower than the dose of 0.6 rad cited above. Also see the response to comment 217.

246. A growing literature suggests that older adults are more sensitive than younger adults to carcinogenic effects of ionizing radiation. This may be due to the accumulation of genetic "hits' from previous exposures, functional declines DNA repair processes, and lowered immune function. The potential for tritium releases to disproportionately impact older adults, as well as populations with other inherited or acquired forms of heightened susceptibility, is an important public health issue that is not addressed in the ATSDR report.

ATSDR Response: The literature concerning people that may be especially susceptible to the health effects of ionizing radiation is extensively reviewed in the ATSDR Toxicological Profile on Ionizing Radiation. While there are no specific conclusions regarding any increased cancer susceptibility of older adults, Table 3-8 includes a statement that "Women irradiated at <= 20 years of age are at higher risk than those irradiated later in life." Also the section 3.2.3 on Carcinogenic Effects from Ionizing Radiation Exposure includes the following statements, "Radiation-induced cancers are the same types that are normally found in an unexposed individual. However, after exposure to radiation, these cancer types may occur with some increasing frequency and therefore can be detected only by epidemiological means. Most of these cancers occur only when those individuals reach an age when these cancers would normally be expected to develop (except for leukemia)."

While we have not repeated that entire review in this health assessment, we have relied on the MRL that was developed using that information. The derived MRL is protective of those people who may be especially sensitive to the effects of ionizing radiation.

247. Basic epidemiologic concepts are poorly applied in the ATSDR report. First, as noted above in discussion of the A-bomb survivors, the ability of epidemiologic studies to detect low-level effects is often compromised by problems of poor measurement, confounding, and inadequate sample size (10). No evidence of an effect is not the same as evidence of no effect. Many studies cannot detect effects because of inadequate dose information, reliance on cause of death information rather than incidence, selective exposure of individuals with lower susceptibility, and lack of long-term follow-up of large populations. The studies cited in the report section entitled "Health outcome data evaluation" (pp26-28) suffer from these problems and should not be interpreted as providing evidence of no effect. Furthermore, ATSDR's assumption that "tritium doses from the accidental LLNL tritium releases are below levels expected to produce adverse health outcomes" (p 26) suggests that any positive findings from occupational and community studies would not be interpreted by ATSDR as evidence of health effects due to their a priori assumption that doses and dose-response relationships are too small for effects to occur.

ATSDR Response: Please see the response to comment 241. Also, it must be noted that scientific observation never does a good job of proving a negative and epidemiological studies have a number of limitations. However, our determination of "no apparent public health hazard" is based on well established public health practice and a weight of evidence approach to the evaluation of low dose radiological exposures. This evaluation, which includes numerous health protective assumptions, is protective of the public health of the Livermore community.

248. Another basic epidemiologic concept is that small relative risks can have a large impact if many people are exposed. Small doses to large numbers of people further away from LLNL could produce a larger health impact than higher doses to small numbers of people near the site.

ATSDR Response: Please see the responses to comments 211 and 227.

249. ATSDR has conducted a biased review of evidence for health effects of ionizing radiation in general and tritium in particular. Important scientific and public health considerations are omitted. The report should be revised to reflect the substantial uncertainties in the scientific basis for understanding the full range of biological effects of low level ionizing radiation and the existence of a growing body of evidence that radiation health effects have been underestimated in the past.

ATSDR Response: Please see the response to comment 237.

Next Section     Table of Contents