Summary: This review finds that the final report is self contradictory in several places, fails to discuss several factors which make its estimates unrealistic, and uses scenarios which are historically impossible. These attributes and a failure to explain to the lay public the uncertainty in the uncertainty estimates give an exaggerated perception of risk to public health. The lay reader is advised not to jump to conclusions. The estimated risks are not realistic and do not express the hazard to the public.

Background: The ORHASP Studies were carried out by a state appointed Panel and their contractors under a $15 million grant from DOE. The work extended over a period of about eight years. Although the meetings were open to the public, the Final Report was not available for review. However many of the errors referred to below were brought to the panels attention  by the public and technical reviews as were other errors. Some but not all errors were corrected. The general overall conservative bias and consequent estimated risks were not known prior to publication.

This study falls on a community, Scarboro, some of whose residents are already laboring under the impression that their community has been seriously contaminated although the sampling and monitoring data would indicate otherwise. The tone of this report is most unfortunate.  It is also unfortunate that there was little attempt to reconcile the conclusions of this report with the known and readily available data on the current levels of these contaminants in sediments, soils and people.

This review is intended for the lay audience. It is assumed that the typical lay reader will not study the several supporting technical reports and will draw their conclusions from the final report alone. Unfortunately, the topic is technical and not all technicalities can be avoided. The review, beyond a few obvious comments, of the several supporting technical documents is not attempted and is left to others. The review tries to place dose reconstruction results in a real life perspective. It is of the utmost importance that the readers of a risk assessment understand, not only the results, but also the assumptions, which were made in reaching those results. The devil is often in the details. The reader needs to be aware that risk assessment, its models, its data and its objectives are highly controversial in the technical community. While most of this review is based on the Final Report, there are a few comments based on the  technical reports.

Objectives of Risk Assessment

In order to interpret the results of a risk assessment with respect to the real risks to an individual, one must realize that risk assessments are carried out to meet different objectives. There are two broad categories of risk assessment based on objectives:

1. Reality-based risk assessment.

The objective of this risk assessment is to estimate actual risk to real populations. This risk assessment attempts to provide an unbiased estimate of the probability distribution function (or its characterizing parameters) of the risk resulting from a specified, realistic, risk exposure scenario. The objective is to estimate a level of risk such that an actually exposed population would show observable evidence of the estimated level of risk if actual exposure to the scenario occurred. This objective is difficult to reach in the absence of unbiased human or species-specific data. In actual practice only the sketchiest data may be available for many types of human risk.

One ignores this type of risk assessment at their peril unless one is certain that they are in the sub-population that has a high tolerance for the scenario's risk factor.

2. Regulatory-based risk assessment.

The objective of this risk assessment is to estimate a level of risk (or contaminant) that will be safe for a human population. This risk assessment attempts to provide a biased estimate of the probability distribution function (or, more frequently, a point value) of the risk resulting from a regulatory-specified, conservative, risk exposure scenario using regulatory-approved parameters. The objective is to estimate a level of risk (necessarily biased) such that any site remediation goal based on the risk assessment will be unquestionably protective of the health of the most sensitive sub-population regardless of cost. This objective is easily met by adopting conservative estimates (often specified by regulation) whenever there is any doubt about an unbiased estimate of any species-specific parameter.

The interpretation of this type of risk assessment is subject to the evaluation of the several conservative estimates of parameters and the over-all degree of conservatism in the estimate. In actual practice the degree of conservatism is dependent upon the availability of human data in the exposure range of the scenario. Unfortunately, very little accurate data is available. Most parameters are determined from animal data or by long extrapolations that are debatable. (See Appendix A.)

Exceeding a safe level, which has a large safety factor is not necessarily risky; it may be another safe level with a smaller but still adequate safety factor. Safety factors such as described in Appendix A or low levels of extrapolated risk make it very difficult to relate regulatory-based risk to reality-based risk. However, it seems safe to say that if the regulatory risk is no greater than ten times the proposed safe regulatory level there is no cause for alarm. A level of 100 times the proposed safe level would be reason for further study, including the safety factors. As an example, the original proposed remediation goal for mercury in the LEFPC flood plain was increased by a factor of forty without encountering a significant risk to the public.

It is of paramount importance in applying risk assessments to real life situations to be fully aware of the objectives of the assessment and the assumptions actually made by its originators. One also must consider the level of the estimated risk compared to other risks in their lives. The importance of risks very much smaller than other common risks is very subjective and driven more by perception than reason.

ORHASP Risk Scenarios - Real or Science Fiction?

The recently released ORHASP Dose Reconstruction Report posits eight dose scenarios claimed to estimate risk "as realistically as possible" (see page 23). However, a casual reading of them raises several questions as described below.

Scenario #1 - A Woman Raised on an East Fork Poplar Creek Farm

"In this scenario, a woman is born in 1940, in 1943, she moved to a farm on the banks of East Fork Poplar Creek in Oak Ridge. She drank backyard cow's milk as an infant and adult. …"

History - In late 1942, the Army took possession of all the land along EFPC. It was not until about 1958 that any of the land was made available as farms to private owners and the homes were built on six farms around 1960 and later. The releases of I-131 stopped in about 1956 and the cows raised on LEFPC were Black Angus and Charlais beef cattle, not milk cows.

Update (3/25/00): Bob Peele has call my attention to a farm family who lived from 1945 to 1957 on a hill on the south side of LEFPC in the vicinity the West End Fire Station. The farm, rented from AEC,  consisted of 200 acres and appears to have been above the flood plain in the area of upper Wiltshire Drive.  The interviewee reports not fishing  but the children did swim did swim. The interviewee also identified another family which lived on what is now the Shannon property during the same period and who also raised cattle. It should be noted that the departure date precludes exposure to mercury.

Scenario #3 - A Woman Raised in the Scarboro Community

"This scenario involves a woman born in 1948 in what is now the Woodland community. She moved to Scarboro in 1953 and continues to live there. She played in EFPC almost every day in the summer from age 5 to 10 (i.e., from 1953 to 1958). …"

Geography - The EFPC is about 3330 feet from Scarboro as the crow flies and about 4400 feet down grade along the tributary. How many parents allow 50 to 10 year-old girls to wander that far every day all summer long?

Scenario #5 - A Man Raised near the Oak Ridge Swimming Pool

"In this scenario, a man born in 1943 moved in 1943 to a location near the Oak Ridge swimming pool. … In 1954 he shot a deer that was contaminated with several radionuclides, primarily Cs-137, released to White Oak Creek and he and members of his family consumed it. …"

History - There were few or no deer on the ORR until the 70's and the hunting started in 1985. Further all legally killed deer were examined and contaminated deer were destroyed.

Scenario #8- A Woman Born in Oak Ridge In 1960

"This scenario concerns a woman born in 1960 who has lived on Lower Louisiana Avenue in Oak Ridge her entire life. She drank backyard cow's milk as an infant and as an adult. …"

Geography - Most lots on Lower Louisiana Avenue are well under an acre and their backyards are even smaller. It takes over an acre to support one cattle unit in East Tennessee and the zoning laws prohibit raising cattle in residential neighborhoods. The text of the report says "There were probably no dairy animals at any of those urban locations." One must ask, 'Why does the report describe a scenario that it also rejects?'

Granted the above scenarios are hypotheticals but what purpose does an impossible hypothetical serve?
The remaining four scenarios are at least possible but whether they represent the life styles of significant sub-populations so as to inform the general public of their risks is debatable.

All of these scenarios deal with Iodine-131, Cesium-137 (or equivalent), PCBs, or mercury in one of several forms, organic, inorganic or elemental. The exposure levels deemed harmful are based on regulatory parameters and suffer from their conservative factors. The results are predominately presented in graphic form but the qualifying remarks are found in the text and sometimes in appendices. This graphic presentation tends to emphasize the upper limit of the risk.  For instance:

• The text mentions that the I-131 dose response curve is still controversial and the uncertainty range in the risk is about three orders of magnitude.
• The most probable risk of the C-137 is below 10**-4 for all scenarios and the upper 95 % confidence limit slightly above 10**-4 for only two scenarios.
• The recent ATSDR Watts Bar Reservoir PCB and mercury in blood serum of heavy fish eaters, showing no elevated levels was placed in an Appendix. Inorganic mercury was calculated as the soluble mercuric chloride not the existing insoluble form, presumably mercuric sulfide.
•  Mercury exceeded the RfD in only one of the "impossible" scenarios".
• In general the exculpatory evidence is not readily accessible to the lay reader.

This section cites specific portions of the text in bold-faced type followed by comments The reader should read the reference and surrounding text to get the full sense if the comment.

General Comment – There are several places where exposures or short duration occur or are implied to occur but there is no distinction whether the dose is considered acute or chronic. These comments have assumed the chronic RfD or lifetime risks are implied.

P.6 "These results suggest it is likely that some people were hurt by the releases."

The lack of quantification in these statement leave it up to the imagination of the reader what the probabilities, number of people and extent of harm are.

P. 7 " Levels of exposure were calculated considering the lifestyles of those people most likely to be exposed to the release."

There is no estimate as to the number of people who had the lifestyle of the hypothetical scenarios and as pointed out above some scenarios were impossible.

P. 9 " … were exposed to some toxicant at a level of 0.2 to 20 micrograms per kilogram body weight. This would mean the scientists believe with a 95 percent confidence that the dose was greater than 0.2 but less than 20 micrograms per kilogram."

There is nothing sacred or universal about a 95 percent confidence level. The range could be the upper and lower bounds, the 90 percent confidence range or any other percent unless clearly stated.

P. 10 "Mercury if found in the environment in three general chemical categories: elemental or metallic mercury, inorganic and organic compounds."

It is a technical error to consider the very soluble mercury compounds and the very insoluble compounds as the same category. The gross solubilities of mercuric chloride and mercuric sulfide differ by almost seven orders of magnitudes and their ionization products differ by more than 50 orders of magnitudes. Mercury ion is also rapidly bound to the organic/clay soils of East Tennessee. These differences are reflected in their biological behavior.

P. 11 – "… it appears that East Fork Poplar Creek families … were likely exposed to levels of PCBs which would not be considered acceptable by today's environmental and regulatory standards."

This statement is true only if large quantities of fish were caught and eaten (See scenario #1). Actually almost no fishing is observed on LEFPC in the vicinity of the farms. Certainly large quantities of fish are not caught and the recent remediation project did not consider the observed PCB levels to be a problem.

P. 12 – "Dose reconstruction seeks to estimate a realistic, plausible range of exposure levels for members of the general public."

This certainly is not true for chemical toxicants where the exposure levels are compared to the admittedly conservative RfDs, which are established for the "sensitive sub-population. Whether it is true for carcinogenic effects is debatable. (See also page 23.)

P. 12 – "Because the level I calculations intentionally biased estimates to over predict consequences, materials for which screening indices did not approach the Panel's guides under these conditions were not considered further."

This is not true. Level I Uranium fell below the guide but this was considered counter intuitive and was included in the Phase II studies. (See page 13.)

P. 13 – " The screening studies suggest that lead releases from Y-12 and arsenic leached from the steam plant ash piles at K-25 and Y-12 should receive priority in any future dose reconstruction investigations."

There is no indication of the lead origin except to say that it is "airborne". The failure to describe the release mode, the exposure route and the risk levels for the Level I and II screens makes it very difficult for the reader to evaluate the meaning of these (and later) statements on this topic. The failure to describe the far larger arsenic releases for the neighboring Bull Run and Kingston steam plants does not allow the reader to see this contaminant in perspective.

P. 13 – "Because the second-level screening calculations employed some assumptions that tended to inflate the screening indices, the panel has recommended …"

It is difficult to reconcile an inflated Level II screen with the claims on page 12 that estimates are "realistic". This topic will reappear later in the report.

P. 18 – "The Oak Ridge Reservation: An Historical Perspective."

It would have been nice if the report had included changes in the Oak Ridge housing patterns and locations. Several implausible statements would have been avoided.

P. 21 – "Current doses of each of these contaminants, particularly mercury and PCBs, could be accessed through, hair, and blood samples."

Current doses of mercury and PCBs were measured by the TDH and ATSDR and the levels were found to be normal. These facts were known to the Panel. Why they do not try to relate their predicted results with data is not explained.

P. 23 – "Dose reconstruction seeks to estimate doses of radiation or toxic substances as realistically as possible for a variety of people, taking into account their life styles, diets and places of residence."

This claim also occurred on page 12. Please see that response.

P. 24 – "When the number of persons fitting the description can be determined, the number of persons who received more than the threshold dose can be approximated.

This statement is conditionally true but the number of residents attributed (see page 33) Scarboro (6000 to 10,000) and the LEFPC farms with homes (40-200) is far greater than the residential capacity of the communities. If the numbers are correct, there must have been a large number of transients who would not have received the calculated doses unless this correction was made. Nowhere is this discussed.

P. 28 – "Areas most significantly affected by releases from the 1954 accident were in the west part of Oak Ridge near Wiltshire Drive….

The Wiltshire Drive subdivision was not developed until about 1959 and iodine-131 has an 8-day half-life.

P. 28 – "There were probably no dairy animals at these urban locations."

True. Why then the scary backyard cow scenarios?

P. 29 – "At Bradbury, about 3.7 miles southwest of ORNL, the doses attributable to the ORNL releases are roughly twice those calculated for the NTS releases."

It would have been useful to the reader to compare at this point the total risk of the vastly wider NTS exposures to the local ORNL exposures.

P. 30 – "It is important to note that the epidemiological studies discussed here neither prove or disprove a relationship between iodine-131 exposure and thyroid disease."

This degree of candor about the lack of proof of the dose response curve is not evident in the report sections where the estimates of health risks are discussed.

P. 31 – "These [mercury] releases continue to this day, although the heaviest by far were from 1953 through 1960."

The reader could have been informed that the current release concentration approximates drinking water standards. It is also important to the EFPC scenarios that there was a major shift in housing patterns along LEFPC about 1960.

P. 32 – "Chronic ingestion of the inorganic forms on 0.3 milligrams per kilogram per day can cause kidney damage in rats."

This is true for the very soluble mercuric chloride but the very insoluble mercuric sulfide and the insoluble forms present in LEFPC floodplain soils have been shown to be far less toxic in humans and rats.

P. 33 – " The dose reconstruction team estimated that, altogether, from 6,000 to 10,000 lived in Scarboro between 1950 and 1958 and from 40 to 200 altogether lived on farms along East Fork Poplar Creek."

Based on the housing available, these numbers are possible only if a significant portion of the residents were transients. Under these circumstances, they would not have received the calculated dose. Further, housing construction started on the six LEFPC farms in late 1958 and was not complete until the 1970s.

P. 33 – "Farm family children from 6 months to 3 years likely received doses higher than the inorganic mercury RfD from 1953 to 1958, while adults likely received doses higher than the RfD from 1955 to 1958."

As stated earlier, there were no farm families prior to about 1959 when the first farm houses were constructed.

P. 33 - "Children in Scarboro or elsewhere who played actively in East Fork Poplar Creek likely received doses higher than the RfD from 1955 to 1958."

LEFPC does NOT flow through Scarboro. It lies about 3300 to 4400 feet from the housing. How likely is it that young children played there regularly and received a significant dose in the few short years of exposure.

Although using uncertainty factors of 10, 100 or even 1000 may result in RfDs lower than strictly necessary to protect public health, the values may still sometimes underestimate the actual risk."
This statement, while true in the strictest sense, is exceedingly misleading to the lay public. One can always postulate that there is some unknown, undetected health endpoint that exists at some fantastically low dose. However the probability is very, very small and until that unknown is discovered one should not assume that risk exists.

P. 34 – LOAELs, NOAELs and RfDs – "It is important to point out that calculated levels of exposure such as those presented in these studies are not risk estimates."

This statement is perhaps the crux of the entire study: The authors do not believe that what they have repeatedly represented as risks throughout the reports are really "risk estimates". It is rare that a report impugns itself.

P. 34 – "The resulting large uncertainties led to large reconstructed methylmercury dose estimates even before 1950, before the Y-12 mercury inventory was significant."

One must question a methodology that leads to large doses before the contaminant is present to be released.

P. 36 – "Because human cancer risk estimates for PCBs are extrapolated from animal studies in a way that likely overestimate risk to humans, the estimated risks to humans are likely too high."

This statement (which is correct) is incompatible with the earlier claims the risk estimates are "realistic" (see page 12 and 23). An estimate cannot be "realistic" and "too high" at the same time.

P. 37 – "(See page 69 for information concerning ATSDR's 1997 study known to eat significant quantities of fish taken from Watts Bar.)"

Of the many references to the risks of eating fish contaminated with PCB's, this is the only reference to the ATSDR finding that elevated levels of PCBs (and mercury) were not found in heavy fish eaters. This is just one example where exculpatory evidence is not prominently displayed or ignored. It cannot be claimed that the time difference is significant for it is claimed that the release of PCBs continues to the present.

P. 51 – Screening Methods - This section describes the three screening procedures as follows:
 

* All cases were referred to the conservative EPA RfDs and slope factors.

It requires very careful reading to ascertain which cases had what degree of conservatism. Please see comments on individual scenarios and computations for each contaminant. Further when Level I screen for uranium was found to be less than the cutoff, the work was continued into level II (see pages 12 and 13).

Pages 55 and 56 present results of several elements whose Level 1 screening were above the Panel's guide but whose Refined Level 1 results were below the guide. Still these elements are recommended for further study albeit at lower priority. One must ask; when do the results of successively improved screens become definitive? Will being above the level 1 screen lead to low priority studies forever?

P. 52 – "In the absence of more appropriate soil samples, the uranium concentration in a sediment sample from East Fork Poplar Creek was used in the screening calculation even though these samples were also obtained in relatively recent years. Estimates of the soil concentrations from the uranium soil in Scarboro (P Volilleque, 1998 [incomplete reference]) suggest that the sediment sample used could have been ten times higher than the peak concentration in Scarboro soil."

Why the substitution of a sample, from a sediment as opposed to a soil sample from a location at least 3000 feet from the estimate location, is indeed a strange approximation especially when the soil and sediments sample along LEFPC vary by order of magnitudes. The fact the uranium in Scarboro was deposited by an airborne mechanism and the sediment sample was water borne makes it even stranger. How was the "correct" sediment sample chosen? Surely these results should be reduced by a factor of ten or based on the 1983, 84, 85 and 1998 soil background levels for Scarboro.

P. 55 – "Using these new estimates of total [uranium] releases, the Phase II investigators recalculated the potential health risks of nearby residents using the conservative Level I screening methods."

The reader must be careful not to associate all phase II calculations with the "best" estimates. When the refined Level I screening was performed the exposure level dropped by an order of magnitude (see page 55, p.4).

P. 56 – "The cancer indices for arsenic were estimated [at K-25 and Y-12] to be three to nine times the Panels decision guide. These releases were associated with the coal-burning process steam facilities at each of the plants."

There is no detailed statement of the exposure route for arsenic although in another section (see page 13) mention is made of "leaching from steam plant ash piles". Again, a comparison of these releases to the releases of the TVA steam power plants would have been useful.

P. 56 – "Automobile exhaust was a large source of lead in the environment in the years when the Y-12 releases occurred. The Y-12 releases would have been in addition to those produced that produced internal contamination that now would be considered to be unacceptable."

It is not clear what the source of this airborne lead is and how its magnitude relates to the automobile source prevalent both at Y-12 and off site.

P. 57 – ORHASP Decision Guides

This limited discussion comes after extensive use of the guides in earlier sections. Nor does it contain any discussion of the many factors that make these estimates regulatory-based rather than reality-based as claimed. One cannot expect the lay public to be fully acquainted with all the technical and bureaucratic limitations that have been placed on risk assessment. There is no discussion of how lifetime exposure guides have been applied to exposures of limited duration.

P. 58 – Uncertainties in Dose and Risk Estimates

This section is inadequate to present to the lay public the intricacies of the interaction of uncertainty of knowledge with the innate variability of a population. Specifically, how this ignorance results in inflated confidence ranges and unreliable confidence limits. The reader has no indication in the uncertainty of the uncertainties. This can be seen in the extremely wide range in the risk distribution, especially for I-131.
This combined with a tendency to dwell on 95 percent confidence limits as well as the graphic presentations, results in a distorted view of the probable actual risks to the public.

P. 61 – "The presence of this nuclide in excess of its natural abundance is a clear indication that at least some enriched uranium releases from Y-12 were carried from Scarboro during some of these years."

This statement, while undoubtedly true, misses the point, which is how much of the uranium in Scarboro soils is a result of this transport. The increase in the enrichment ratio of the current Scarboro is small and the total uranium is within background range. East Tennessee clay soils adsorb uranium. From a simple mass balance making simple assumptions, it is possible to calculate the possible contribution of Y-12 to the Scarboro uranium. It varies from insignificant (0.2 %) of back ground to about 21 % of background depending on method and assumptions. The best estimate based on the ORHASP estimates of U238 and U234/235, is about 6% for the residential part of Scarboro and 20 % on Pine Ridge. None of these emissions  have health significance as the levels are at background.

Further, the Y-12 wind-rose pattern is never discussed. It shows about 5 % directed toward Scarboro and the remainder to the east or west. If Scarboro has received significant amounts from Y-12 there should be a significant soil-uranium footprint in the Y-12 valley to the east or west approximately 10 times the alleged Scarboro footprint. There is not.

P. 79 External Technical Peer Reviews of Phase II Documents

This section does not give any indication of the completeness of the reviews nor of the dissenting comments of the reviewers. Nor is the list of reviewers and commenting stakeholders complete.  A very insightful review by George D. Kerr of  ORNL (letter to Timothy Joseph of DOE) is not included. The review by Auxier and Associates is also revealing. It is not clear if the reviewer's comments will be made available.

Reviewers of earlier versions expressed many other technical concerns about the conservative nature of the ORHASP dose reconstruction. It will take considerable time for expert reviewers to study all of the final detailed reports but if the validity of the above scenarios is an example of the reality of other assumptions made, there will be much room for criticism.

Task 2 – Mercury Releases from the Y-12 Lithium Enrichment

1) ES 5 p4 – The EFPC Floodplain Family …. This is not a hypothetical exposure scenario –--… The total size of the population between 1950 1n 1990 was estimated to be between 40 and 200.
Comment: There were six farmhouses built between about 1960 and 1975. How can this not be  hypothetical scenario as  claimed in the panel report. (see earlier discussion of Scenario #1.)

Task 6 – Uranium Releases from the ORR ….

1) Page ES-11 p1 – Late in the project, new estimates indicated that the Y-12 emissions for 1986 – 1995, the calibration years, were low. This would lead to high predictions of emissions for the early years. The new estimates were not used leaving a conservative bias.

1) Page 3-19 P1 – The upper 95% confidence level of the X/Q values was used rather than the average. This is a conservative factor of 1.5.

The following is a brief summary of the major sources of conservatism of regulatory risk assessment. If all are applied the overall safety factor can be as high as one million although the usual intent is to have a safety factor in the range of 100 to 1000. There are four broad categories of conservatism in regulatory risk assessment based on the model, type of risk and source of data:

1. Scenario Parameters

One must evaluate the risk scenario in terms of how well it fits the actual real life exposure to the risk. Often there is the assumption of very improbable land uses or improbable intensities of reasonable uses and the real risk must be judged accordingly. Scenarios also include estimates of doses from specific exposure routes, which may be exaggerated.

Dose reconstruction scenarios often suffer from a paucity of release data and the errors of the contaminant-dispersing model. Biases are difficult to avoid. If the methodology attempts to evaluate the unknown uncertainties in the model parameters this has the effect of increasing the 95 percent confidence range of the results. This makes results appear more risky than they really are. Dose reconstruction often relies on conservative regulatory procedures to convert dose to risk.

2. Toxic Effects based on Animal Data

This usually involves using toxicity data (No Observed Adverse Effect Level) from a sensitive strain of a sensitive animal species. Followed by the assumptions that humans are ten times as sensitive as the test animal and sensitive humans are ten times as sensitive as the average human. In the absence of NOAEL data the Lowest Observed Adverse Effect Level (LOAEL) may be reduced by a factor of ten to estimate the NOAEL. In some cases the Averse Effect of the LOAEL is not serious and is reversible. It is well known that humans are more tolerant of some toxins than some species of test animals and that some toxins affect animals by mechanisms not significant to humans. Sensitive populations at times are only two times more susceptible to the toxin than the average human. At times data for a soluble chemical form is substituted for the missing data of an insoluble chemical form.

If all of these factors and assumptions are applied then the safety factor (often called uncertainty) may be as high as a million but factors of 100 to 1000 are more common. Few regulators would be happy with a safety factor as small as 10 else the remediation goals could be questioned.

3. Toxic Effects based on Human Data

Human data is sometimes available from occupational or accidental scenarios but the actual dose may be in question. Further the study must be corrected for confounding factors or some other effect may be measured. For some less serious Averse Effects some controlled exposure data is available. Having human data allows the safety factor to be reduced to a factor of 10 to 100 under many circumstances but great care must be observed in its use.

4. Carcinogenic Effects of Chemicals from Animal Data

Carcinogenic effects are assumed to be proportional to dose and the dose-response line is extrapolated from high doses where effects are observable to zero dose where effects are not observable. If there exists, as some authorities claim, a non-zero threshold below which the cell repair procedures overcome the damage then the linear extrapolation is extremely conservative below this threshold. Further, the practice is to use the upper 90% confidence of the empirical slope, which in itself is conservative. (See the discussion in the next section.)

5. Carcinogenic Effects from Radioactivity

Carcinogenic risks are based on human data from the Hiroshima atom bomb studies. Unfortunately this data is at a dose rate about 10**15 times the dose rate of a modest environmental exposure rate. The spontaneous thermal (chemical) damage to a single mammalian cell is estimated to be 70 million per year while the background radiation damage is only five. There is no guarantee that the mechanisms for cell damage and repair are the same with such drastically different dose rates and dose levels and there is most certainly some error correction mechanism. The regulatory practice is to extrapolate in a straight line from high dose rates to a zero dose rate without any threshold level. There is a significant amount of data that indicated a threshold does exist and that there is even some benefit accrues from low doses. There is increasing evidence that cell damage mechanisms exist and low exposures activate the immune system against later larger exposures. The conservative regulatory practice ignores these arguments and the more recent low-rate exposure data.

Further, the risk goals are usually set in the range of 10**-6 to 10**-4 for the lifetime cancer risk. These are a factor of 1000 to 100,000 less than the normal lifetime cancer risk of about 3 out of 10. These are small risks, which could not possibly be detected by an epidemiological study. Applied to a neighborhood of 1000 people these risks would account for less than 0.1 to 0.001 deaths per seventy years.

6. Multiple Biases or Extended Distribution Functions

When multiple conservative biases are applied to the computation of a risk, the result is cumulative and very large, conservative overall safety factors can arise. This is especially true when the individual biases are a factor of ten and sometimes more.

In a similar manner, broadening a distribution function to account for lack of information on the range of a statistical variable will broaden the resulting risk distribution function, giving rise to conservative confidence levels of the predicted risk. Changing a triangular distribution function to a uniform distribution function has the same effect. Repeated application of this leads to compounded results. Even the substitution of the physically impossible normal distribution, with its infinite tails, can lead to inflated risks.

None of the above should be interpreted to mean that risk assessment (or dose reconstruction) is not a valid tool but only that the reader needs to be aware of all the details and assumptions before deciding the results are real. This is especially true if the risk assessment has been guided by regulatory concerns whose objectives are admittedly to make safe, but not real, predictions.