This closing statement is arranged in five parts:
Method for Handling the Retroactive Alteration of Dose Estimates in the A-Bomb Study, p.1 Cancer-Risk at Moderate and High Dose-Levels, Acute Delivery Only, p.3 Cancer-Risk at Low Doses, Acutely and Slowly Delivered, p.7 Disproof of Any Safe Dose or Dose-Rate, p.13 Some Practical Implications for Human Health, p.15
At the close of this book, we must return to the question posed by the book's title, and by Chapter 2: Do conclusions in this field come out differently from an independent analysis, than from analyses provided by the radiation community?
The answer is yes. This book has examined five main topics, listed above. (Hormetic speculations are in a class by themselves, and are examined in Chapter 35.) Except on the least important topic -- the cancer-risk from acute exposure at moderate and high dose-levels -- we differ seriously with most of the radiation community.
Moreover, we are severely critical of scientifically questionable practices and apparent inconsistencies which exist in some other analyses, and which we have identified in the preceding chapters.
In this closing chapter, we shall compare our own conclusions on each of the five main topics with those of most of the radiation community, and in particular, with those of the 1988 UNSCEAR and 1990 BEIR-5 Committees.1. Method for Handling the Retroactive
Alteration of Dose Estimates
in the A-Bomb Study
Uncertainties remain in the field of radiation epidemiology, and the database which is most capable of resolving them in the future is the A-Bomb Study. Thus humankind as a whole has a stake in protecting this uniquely valuable study from practices which can destroy its credibility in the future.
We have shown in Chapter 5 that a process is now underway of substituting a retroactively altered database for the study's previous database. This is not planned as an addition of new dose-estimates to the study. What is occurring is replacement of the previous dose-estimates and even replacement of the entire structure of the A-Bomb Study.
Everyone welcomes possible new insights about dosimetry in the A-Bomb Study, but there are right ways and wrong ways to add new information to an on-going prospective study. It is essential to show the scientific community that introduction of new information is not creating opportunities for intentional or unintentional bias to enter a revised study, for such opportunities by themselves are sufficient basis for the scientific community to reject a study as unreliable.
Introduction of the new DS86 dosimetry has placed the A-Bomb Study at a crossroad. In Chapter 5, we have shown -- using the very principles of epidemiological science -- that the A-Bomb Study's foundation as a legitimate prospective study will sink into quicksand, unless a way for handling the new dosimetry is developed which restores the study's continuity and maintains the continuity during the study's remaining decades of follow-up.
The farthest thing from the truth would be for anyone to say that we are objecting to possible new insights about dosimetry in the A-Bomb Study. (Indeed, the new DS86 dosimetry confirms the correctness of our own handling, almost ten years ago, of the mistaken neutron dose-estimates.) Far from objecting, we are just determined that improvements in dosimetry be handled in a way which will not ever impair the credibility of this never-to-be-repeated study.
-- Part 1, Our Finding :
This book not only proposes, but also demonstrates, a practical "constant-cohort, dual-dosimetry" method for having the benefit of possible new insights about dosimetry, while maintaining the continuity and identity of the A-Bomb Study as a first-class prospective study with a permanent structure.
Our method (described in Chapter 6) keeps the cohorts of survivors together, exactly as they were before the new DS86 dosimetry. For each cohort, we just calculate a DS86 dose-estimate which is additional to the cohort's previous T65DR dose-estimate, and then we analyse cancer-hazard in both the old and new dosimetries for the same sets of people.
In this way, scientists can have the potential benefits of the new DS86 dosimetry, but DS86 never replaces the previous dose-estimates and never disturbs the study's permanent prospective structure. The unaltered database would remain, forever, the stable and objective foundation ensuring the scientific credibility of the whole study. With our method, there would be no retroactive changes of input at all -- only some supplemental information about doses. Therefore, the "constant-cohort, dual-dosimetry" approach protects the A-Bomb Study from rejection based on concern that the current version of the new dosimetry and its future revisions might be opportunities for bias to enter.
A Warning from "Lizzie Borden" :
A major purpose of this book is to help redirect the handling of the A-Bomb Study along sound epidemiological lines. We wish to emphasize the importance of this issue for science in general.
Poor choices regarding a world-famous database would not only set back this field for decades, but would influence work in other fields, too. If it were ever to become universal practice to alter a prospective study's original input after any of the outcome is known, never to reveal what the study would have shown with its unaltered input, and to commit mayhem on the study's original architecture and cohorts, then epidemiology as a credible science would be finished. Some worst-case scenarios were described in our Chapter 5, Part 2. Surely we are not alone in wanting to avert "Lizzie Borden Methods" of database management:
"Lizzie Borden took an axe
And gave her mother forty whacks.
When she saw what she had done,
She gave her father forty-one." (Anonymous.)
Leading figures in the radiation community acknowledge that the retroactive alterations of the A-Bomb Study, now underway, are massive. We would have thought that this obvious challenge to the most basic rules of prospective research would merit thoughtful discussion and remedial proposals by the radiation committees. Nothing of the sort has occurred.
-- Part 1, Radiation Committees :
UNSCEAR 1988 :
There is no discussion of this issue, nothing to quote, except phrases which imply full approval of what is going on:
For instance, the DS86 dosimetry is described in the executive summary as "a revised dosimetric system for the survivors of Hiroshima and Nagasaki that allows a better analysis of this important epidemiological series" (Un88, p.32, Para.192).
Later, the word "replace" appears: "It is particularly propitious that risk assessments be made now, in light of the revised dosimetric evaluations of the Japanese survivors of the atomic bombings, the most important study population. This re-evaluation, completed in 1986 and known as the dosimetry system 1986 (DS86), replaces the previous estimates of 1965 (T65)" (Un88, p.407, Para.3).
BEIR-5 Committee :
Like the UNSCEAR Committee, the BEIR-5 Committee appears ready to replace the T65DR dosimetry by the new DS86 dosimetry, and the report does not show results for the T65DR cohorts or dosimetry.
"The analyses of radiation effects among the Japanese A-bomb survivors in this report make use of new dose estimates developed in a five-year study by Japanese and American scientists. This binational study resulted in a new dosimetry system, designated DS86, which is documented in two recent Radiation Effects Research Foundation (RERF) reports ..." "(Beir90, p.190).
"As the aim of this report is to provide risk estimates based on the best available data, the committee confined itself to analyses using just the DS86 data" (Beir90, p.198).
-- Part 1, Discussion :
We note that both UNSCEAR-88 and BEIR-5 may leave the impression on their readers that retroactive alterations are limited to new dose-estimates. Even if that were the situation, it would not justify sending the T65DR dose-estimates to oblivion. A "dual-dosimetry" analysis would be required in order to maintain the study's legitimate prospective status, with the new dose-estimates handled as a supplement -- not a replacement.
And, unfortunately, the retroactive alterations now underway are certainly not limited to new dose-estimates. We have shown in detail (Chapter 5, and Tables 10-A,B, and Tables 26-N,O) that all the former cohorts are destroyed, and participants are shuffled into new groupings with new distribution of the previous cancer-deaths and new mean ages at the time of bombing and new male-female ratios. Without a "constant-cohort" approach, continuity is gone.
In effect, the prospective study was terminated with the 1982 follow-up, its database was marked for oblivion, and with 37 years of results in hand, a new database was created.
It is hard to imagine a more "questionable practice" in epidemiological research than this. Thus I find it amazing, particularly in a report which carries the imprint of the National Academy of Sciences, National Research Council, that this treatment of a database merits no discussion at all.
By contrast, we say that such handling will unnecessarily send the A-Bomb Study's scientific believability into oblivion -- and we say "unnecessarily" because Chapters 10 through 17 of this book demonstrate a "constant-cohort, dual-dosimetry" approach to the current and future follow-ups which can easily maintain the study's status as an objective, first-rate prospective inquiry.
The "constant-cohort, dual-dosimetry" approach offers everything to gain and nothing to lose.
In Part 3 of this chapter, readers will see for themselves the urgency of a "constant-cohort, dual dosimetry" analysis. Without it, there is trouble even sooner than we expected.2. Cancer-Risk at Moderate
and High Dose-Levels, Acute Delivery Only
Since this is a book about radiation-induced cancer from low-dose exposure, only one of our tables includes any estimate of Lifetime Fatal Cancer-Yield from moderate or high dose-levels. But for the purpose of comparing our work with the new estimates from the radiation committees, we will make the necessary estimates right here.
Readers will see for themselves that there is substantial agreement between the analysis in this book and the 1988 UNSCEAR and 1990 BEIR-5 Reports with respect to cancer-risk per rad from moderate and high doses acutely delivered -- now that those committees have greatly increased their past estimates. On the other hand, we can only compare apples with oranges, as readers will see below.
Preparation of Estimates :
First we can look at Table 13-A, where Row 10 shows that the average internal organ-dose for all the exposed A-bomb survivors combined (Reference Group excluded, of course) was 41.7 cSv in the T65DR dosimetry, and 47.4 cSv in the DS86 dosimetry.
Then we can look at Table 13-B, which includes Lifetime Fatal Cancer-Yields based on all the exposed Dose-Groups (3,4,5,6,7,8) combined. For the T65DR dosimetry, Column D shows the Cancer-Yield to be 7.37 radiation-induced cancers per 10,000 persons, per cSv (or, 7.37 cancers x 10^-4 person^-1 x rem^-1). For the DS86 dosimetry, Column J shows 6.50 as the Cancer-Yield.
We must make two adjustments, in order to make these values properly comparable with the UNSCEAR-88 and the BEIR-5 values.
First Adjustment :
The values in Table 13-B do not take account of the fact that cancers in future follow-ups will be arising in the A-Bomb Study from the most radio-sensitive age-groups. The values in Table 16-B do take this important fact into account. The impact can be measured by the ratio of low-dose Cancer-Yields from Table 16-B over the low-dose Cancer-Yields from Table 14-C.
Table 16-B, T65DR: Yield = 31.65 Table 14-C, Col.C, Row 2: Yield = 12.90 Ratio = 2.45 Table 16-B, DS86: Yield = 30.43 Table 14-C, Col.E, Row 2: Yield = 12.03 Ratio = 2.53
Therefore, to take account of the radio-sensitivity, we should raise the Cancer-Yields for the combined Dose-Groups by these ratios.
acute gamma-ray doses of moderate level,
LIFETIME FATAL CANCER-YIELD (excl. leuk.) :
T65DR: 7.37 x 2.45 = 18.06
DS86: 6.50 x 2.53 = 16.44
This is all we need to make the comparison with the Cancer-Yield in UNSCEAR-88. For the comparison with BEIR-5, we need to adjust our Japanese value to a United States value.
Second Adjustment :
To adjust the two values above for a United States' population, we shall use the ratio of our Lifetime Fatal Cancer-Yields for a U.S. population in Table 16-C, over the Cancer-Yields for the A-Bomb Survivors in Table 16-B:
UNITED STATES POPULATION,
acute gamma-ray doses of moderate level,
LIFETIME FATAL CANCER-YIELD (excl. leuk.) :
T65DR: (26.64 / 31.65) x (18.06) = 15.20
DS86: (25.56 / 30.43) x (16.44) = 13.80
-- Part 2, Our Finding :
We will assemble our findings from above.
LIFETIME FATAL CANCER-YIELD excluding leukemia.
Radiation-induced fatal cancers per 10,000 persons, per rad, from a single, acute gamma-ray dose of moderate level :Japanese United States T65DR = 18.06 T65DR = 15.20 DS86 = 16.44 DS86 = 13.80
These Cancer-Yields should be doubled if exposure is from diagnostic X-rays.
-- Part 2, Radiation Committees :
At the end of this section, we will assemble all the values for comparison.
UNSCEAR 1988 :
UNSCEAR-88 says: "The atomic bomb survivors have been used in this report as the main source of risk estimates, while the Committee notes that other sources of data such as the ankylosing spondylitis patients are in general terms consistent with these estimates, especially when the mode of delivery of the exposure is taken into account. The Committee has not itself made primary estimates of risk in the Japanese atomic bomb survivors, but has relied on risk estimates developed in recent publications ..." (Un88, p.490, para.594). Elsewhere, Un88 acknowledges Shi87 and Shi88 (TR-12-87 and TR-5-88) as the source of its risk-coefficients. Only the DS86 dosimetry, and only the abridged cohort of 75,991 survivors, were used.
Values in the UNSCEAR-88 tables do not include the correction-factor of 1.23 for underascertainment of cancer in the A-Bomb Study (see our Chapter 11). UNSCEAR warns its readers (Un88 , p.485, para.557): "... the risk coefficients that are used have been obtained from published reports and do not take into account the underreporting of cancer deaths on death certificates. BEIR III, in its projections, increased these coefficients by 23 % to take account of underreporting. A comparable action here would increase the Committee's projections of excess lifetime mortality by "20-25 %."
In its Table 62 (at page 527), UNSCEAR-88 provides its estimate of Lifetime Fatal Cancer-Yield for a Japanese population of both sexes (50 % male, 50 % female) and all ages, exposed to 100 rads "of organ absorbed dose of low-LET radiation at high dose rate, using age-specific risk coefficients" (from the Japanese A-bomb survivors). Since UNSCEAR-88 states (p.479, para.510) that "linear risk estimates are a reasonable summary of the dose-response," the per-rad value from 100 rads in its Table 62 will be the same at a total acute dose of 42-47 rads -- the dose used in our own estimate above.
In Table 62's multiplicative model, the per-rad value for all malignancies (excluding leukemia) is 9.7 fatal cancers per 10,000 persons per rad. (The value for leukemia alone is 1.) The value of 9.7 needs multiplying by 1.23 for underascertainment.
Thus the UNSCEAR-88 value which compares with our "Japanese" value above is 11.93.
It should be noted that UNSCEAR's previous value -- which even included leukemia -- was 1.0. The new estimate in UNSCEAR-88's Table 62 is more than 12-fold higher than its previous one.
BEIR-5 Committee :
Like UNSCEAR-88, the BEIR-5 Report relies heavily on the A-bomb survivors (abridged DS86 cohort) for arriving at its Cancer-Yields: "... the risk estimates that are presented in this report are derived chiefly (or exclusively) from the Japanese experience" (Beir90, p.218). Indeed, the report acknowledges Dale Preston as its scientific advisor. And practices by RERF analysts which we criticized earlier in this book are carried on in the BEIR-5 Report too.
For instance, in addition to excessive subdivision of data, and replacement of the entire T65DR database, and exclusion of 15,000 of the study's 91,231 survivors, there are even exclusions of data from the abridged DS86 database and there are pre-judgments. A few illustrations are in order here.
Five Years of Follow-Up Discarded :
In BEIR-5, there is pre-judgment of a 10-year minimal latency period with the exception of leukemia and breast-cancer. Cancers occurring before 1956 are just not counted in BEIR-5's "preferred risk models" (page 168). The 1950-1955 follow-up has been discarded. This is an extremely questionable practice, as we noted earlier in the book. UNSCEAR-88 (p.524) also assumes a 10-year "minimum latency," for all malignancies except leukemia.
Discarding the 1950-1955 evidence becomes an even more questionable practice when BEIR-5 (p.274, Fig. 5-13) seems to be claiming a time-dependence for lung-cancer which would imply that the ten-year exclusion may result in missing much of the radiation-induced excess with respect to lung-cancer -- one of the most common of all cancers.
Two Dose-Groups Discarded :
In BEIR-5, there is also pre-judgment that dose-response cannot be supra-linear except at very high doses, where cell-killing is regarded by BEIR-5 as an acceptable explanation. This pre-judgment leads to additional throwing out of evidence : "The RERF data show a tendency toward decreased risk per Gy in the highest dose groups, which may reflect either cell-killing or overestimation of the doses in this group. The committee considered various ways of dealing with this problem, including adding terms to the dose-response part of the model and adjusting the highest doses downward. In the end, it was decided simply to exclude the two highest dose groups" (Beir90, p.199).
We note the subjective view of a supra-linear dose-response: It is called a "problem" to find that risk per rad increases progressively as dose decreases toward lower levels. Supra-linearity was viewed as such a bad problem by BEIR-5 that desperation measures were considered (described in BEIR's statement above) -- including even changing the DS86 doses which had just been newly estimated by a group of presumably objective physicists after almost ten years of study (our Chapter 5). The attitude toward the findings, as they fall out of the data, was such that the only solution for BEIR-5, "in the end," was to get rid of part of the data.
One may ask what the BEIR-5 Committee would have done if these two Dose-Groups had contributed to a concave-upward dose-response. Would the Committee have thrown them out?
We regard the exclusion of two Dose-Groups on the basis cited by BEIR-5 as scientifically unacceptable. If there were a solid basis for regarding the finding of supra-linearity as spurious, it would be scientifically objective to call it a problem. But there is no such basis. On the contrary. The finding has been turning up for a decade (our Chapter 22), and now is turning up in both the T65DR and DS86 dosimetries. BEIR-5's own Table 4D-2 (p.200) again confirms that dose-response is supra-linear for all non-leukemic cancers combined, throughout the dose-range shown there. We shall return to the shape of dose-response in Part 3.
Cancer-Deaths beyond Age 75 Discarded :
We will mention one more example of throwing out data. All cancers occurring after age 75 have been excluded from the BEIR-5 risk analysis. "Records of cancer mortality at attained ages greater than 75 years were omitted because of the lesser reliability of death certificate information in such cases, as outlined in Annex 4D" (Beir90, p.165). In Annex 4F (p.218), a statement suggests that the severity of this problem may be magnified by BEIR's own choice to do site-specific analysis. We cannot be sure: "... in that body of data [from RERF], the accuracy of diagnosis from death certificates declines rather sharply beyond age 75, to the point that little reliance can be placed on the data for specific sites. The Committee has refrained from basing analyses on data that it considers unreliable."
BEIR's exclusion of cancer-deaths after age 75 must already affect two of the five age-bands (see our Table 4-B): The age-band 35-49 years old at the time of the bombings, and the age-band 50-years and older (average = 58.5 years) ATB.
An Urgently Needed Remedy :
These three exclusions (follow-up years, Dose-Groups, cancers occurring beyond an attained age of 75) amount to major retroactive revisions in the study's database, all made with 35 years of follow-up results at hand. Each of the three exclusions has its own impact on risk-estimates and on the curvature of dose-response, but presently, only RERF and the radiation committees know what those impacts are.
The three changes are additional to the replacement of the T65DR database and cohorts by the DS86 database and cohorts, and these additional changes presently represent another assault against the continuity of the A-bomb Study and its credibility as a legitimate prospective inquiry with a permanent structure.
The urgently needed remedy would be similar to the "constant-cohort, dual-dosimetry" approach. If all the data are made available by RERF, in a form which maintains analytical continuity with the 1950-1982 follow-up, then no one should object to supplemental analytical work carefully showing the separate impact of each retroactive exclusion upon risk-estimates and upon curvature of the dose-response. Every type of analysis should be welcome providing nothing interferes with the continuity of the on-going study in its unaltered form.
Cancer-Yield -- BEIR-5 vs. BEIR-3 :
After making its chosen exclusions, BEIR-5 does provide Lifetime Fatal Cancer-Yields based on the abridged DS86 subcohort.
BEIR-5's "preferred model" for estimating Lifetime Fatal Cancer-Yield is based on subdivision of the solid cancers into four classes, with separate equations for each class (pp.168-170). For two of the four classes, namely respiratory and breast cancers, BEIR's equations forecast that the radiation effect will fall with various times after exposure. Of course, different equations are used for different ages at exposure. The BEIR-5 Committee acknowledges that its choices were influenced by its idea of how much the risk should vary between those young at exposure and old at exposure (Beir90, p.203):
"The committee considered a variety of models before selecting the preferred models described in Chapter 4 ... In general, the preferred models fit the data as well as the alternatives and have fewer terms. This was not the sole criterion for model selection. The committee paid particular attention to how risks were proportioned between various age groups."
BEIR-5 does not provide Lifetime Fatal Cancer-Yields explicitly for moderate or high doses. In its Table 4-2 (at page 172), BEIR-5 provides a value for a single, acute whole-body exposure of a mixed-age United States population to 10 rems. However, since BEIR-5 states emphatically that it finds the dose-response to be linear (for instance, p.5, p.200), the per-rem or per-rad value from 10 rems in its Table 4-2 is necessarily its per-rem value for moderate and high acute doses too.
In Table 4-2, the per-rem value, for a population of 50 % males and 50 % females, is 6.95 radiation-induced cancer-deaths (leukemia excluded) per 10,000 persons, per rad. (The value for leukemia alone is 0.95.) The value of 6.95 needs multiplying by 1.23 for underascertainment.
Thus the BEIR-5 value which compares with our "USA" value above is 8.55.
It should be noted that BEIR's previous value -- which excluded leukemia and bone cancer -- was 2.0 (Beir80, page 206, Table V-19). BEIR's new estimate is about 4-fold higher than its previous estimate.
-- Part 2, Discussion :
Now we will assemble all the values from above.
LIFETIME FATAL CANCER-YIELD excluding leukemia.
Radiation-induced fatal cancers per 10,000 persons, per rad, from a single, acute gamma-ray dose of moderate level :
Japanese Pop'n United States Pop'n Gofman est. Gofman est. T65DR = 18.06 T65DR = 15.20 DS86 = 16.44 DS86 = 13.80 UNSCEAR-88 est. BEIR-5 est. DS86 = 11.93 DS86 = 8.55All the Cancer-Yields above should be doubled if exposure is from diagnostic X-rays (see our Chapter 13, Part 4; also Beir90, p.218).
Readers can see for themselves that there is less than a factor of two separating our estimates and those of the radiation committees.
On the other hand, assessment of the agreement must remain approximate, due to the fact that we have necessarily compared apples with oranges. The radiation committees and we are no longer working with the same A-bomb database. We alone are providing estimates which use the unabridged, legitimate prospective database, with its constant cohorts and its objectivity ensured by continuity.
We are eager to acquire the necessary data -- with none pre-discarded -- to extend our "constant-cohort, dual-dosimetry" analysis (which presently covers 1950-1982) to include 1950-1985 and beyond.3. Cancer-Risk at Low Doses, Acutely and Slowly Delivered
Where risk-estimates really matter -- for acute-low and for slow-low exposures, our independent analysis indicates that the radiation committees are underestimating cancer-risk by up to 30-fold.
In other words, although the radiation committees have made progress toward realism by greatly increasing their previous risk-estimates for acute moderate-to-high exposures, we think they still lack realism with respect to acute-low and slow-low exposures. These committees are continuing the practice, documented by our Chapter 22, of rejecting good human evidence pertaining to such estimates, and substituting a preferred hypothesis based on non-human evidence.
The preferred hypothesis is that human dose-response is concave-upward. However, our Figure 14-C clearly shows that upward curvature does not describe the human evidence.
If dose-response really were concave-upward, the cancer-risk per rad would progressively diminish as either dose or dose-rate decreased, due to the diminishing opportunity for inter-track carcinogenesis (Chapter 23).
However, the direct human evidence shows a dose-response relationship which is supra-linear (like Figure 14-A) and supports no expectation of reduced risk per rad from acute-low or slow-low exposures compared with acute-high. UNSCEAR-88 makes a claim that some human evidence supports a dose-rate effect, but we have shown the fallacy of the claim in Chapter 22, Parts 4 and 5.
It should be emphasized, also, that even if human dose-response were linear (Figure 14-B) instead of supra-linear, linearity would not support any expectation of reduced risk per rad from acute-low or slow-low exposures compared with acute-high.
"Q.E.D." Summary :
As we illustrated in Chapter 23, Parts 5 and 7, the following relationships occur between risk and dose.
In the absence of a concave-upward dose-response relationship, there is no reduction in per-rad risk between an acute dose of 160 rads and an acute dose of one rad. Indeed, when dose-response is supra-linear, we illustrated how the per-rad risk increases at acute-low doses compared with 160 rads. Thus, one writes:
-- EQUATION (1) for supra-linear & linear:
Per-rad risk at acute-low => per-rad risk at acute-high.
Virtually no one denies that the contribution to cancer-risk from inter-track carcinogenesis (the quadratic term in a linear-quadratic dose-response equation) is extremely small at acute-low doses. Therefore, removal of opportunities for inter-track carcinogenesis by slow delivery of a low total dose has a negligible effect on per-rad risk. Thus one writes:
-- EQUATION (2) for supra-linear & linear :
Per-rad risk at slow-low = per-rad risk at acute-low.
Then by substitution or direct logic, one writes:
-- EQUATION (3) for supra-linear & linear :
Per-rad risk at slow-low => per-rad risk at acute-high.
In short, in the absence of the concave-upward dose-response relationship, there is no basis whatsoever for expecting per-rad risk of radiation-induced cancer to be lower from acute-low or slow-low exposures than from acute-high.
However, we do not rule out the possibility (as indicated in Chapter 23, Part 6) that per-rad risk may be lower for slow-high than for acute-high. On the other hand, in Chapter 23 we showed that slow delivery of a high total dose may not reduce the per-rad risk at all. It may turn out in the human that inter-track carcinogenesis accounts for only a minor part of the radiation-risk, even at moderate to high acute doses. Pending more evidence on the issue, we regard it as premature for anyone to count on any reduction of risk if a moderate or high total dose is delivered slowly.
Conflict over the Shape :
As shown in Chapters 13 and 27, radiation-induced cancer is now provable in the A-bomb survivors from an internal organ-dose as low as 11 rems (15 rems in DS86). Radiation-induced cancer-risk between 11 rems and zero dose (or 15 rems and zero) is necessarily estimated by interpolation. Thus such estimates depend on the shape of the dose-response curve along which analysts make their interpolation.
We will compare our finding about shape with the reports from the radiation committees.
Our Own Finding of Supra-Linearity :
In Chapters 14, 29, and 30, we showed that that dose-response is supra-linear, with the highest cancer-risk per rad in the lowest dose-range. And this finding is shown to be significant. The 1950-1982 evidence fits a supra-linear relationship significantly better than it fits a linear relationship.
We ask no one to accept, on faith alone, our assertion that this is so. Our finding of supra-linearity has been reached openly, step-by-step, from the unabridged evidence. All the work is checkable and verifiable.
We have stated clearly, in Chapter 14, Part 4, that credible evidence might develop in the future to alter the current finding of supra-linearity. Only time, and objectively conducted studies, will tell. Meanwhile, we note that supra-linearity is the shape which has been showing up for three consecutive follow-ups of the A-bomb survivors (1950-1974, 1950-1978, 1950-1982).
Within the 1950-1982 evidence, the dose-response relationship turns out supra-linear no matter how we approach the data. It turns out supra-linear with the T65DR dose-estimates, and with the DS86 dose-estimates. It turns out supra-linear when we test cancer-deaths per 10,000 initial persons versus dose, and when we test cancer-deaths per 10,000 person-years versus dose. And it turns out supra-linear whether we combine males and females, or test them separately.
Very important is the finding that supra-linearity is the dose-response relationship throughout the dose-range, not just at high doses (see Table 13-B).
A great deal has been written about the alleged necessity of basing low-dose risk-estimates on dose-response curves limited to high-dose observations. Now that radiation-induced cancer is provable in the A-bomb survivors at an internal organ-dose of only 11 rems (or 15 rems in DS86), interpolation of radiation-risk is required only in the short segment of the dose-range between zero rems and 11 (or 15) rems. Any suggestion that low-dose Cancer-Yields must be based on high-dose observations would be plainly inappropriate.
UNSCEAR 1988 on Shape :
UNSCEAR-88, introducing its section on radiation carcinogenesis, states (p.407, para.5): "A main concern that cannot be adequatetly resolved is how to relate the results obtained at high doses and dose rates to the low levels of exposure that may be expected in environmental and routine occupational settings."
The suggestion that estimates still depend on high-dose observations is made again soon thereafter: (Un88, p.416, para.68): "Estimates of low-dose risks based largely on high-dose data must depend heavily on the assumptions about the shape of the dose-response curve and are, of necessity, no better than the model is applicable. Current data suggest that resolution of these difficulties will not be easy, and it seems likely that there will be many site-specific differences."
Earlier in this book, we have criticized the practice of subdividing databases by specific organs, creating horrific small-numbers problems, and then taking the results seriously. It is in the nature of numbers that there will never be a database so large that it cannot be rendered inconclusive and unreliable by excessive subdivision. It should be evident that one can lose the prospect of ever obtaining answers to certain questions simply by excessive subdivision of the data.
What does UNSCEAR-88 say about the shape of dose-response in the A-Bomb Study? The report acknowledges that linearity provides a good fit, and not only for all cancers combined, but even for subdivisions -- which would seem to undermine the suggestion about "many site-specific differences" above. We quote (Un88, p.479, para.510):
"Over the range of doses from 0 to 6 Gy, there is no clearly significant evidence of non-linearity (although other forms of response fit the data), so from a purely statistical point of view linear risk estimates are a reasonable summary of the dose-response. Moreover, when linear, quadratic, and linear-quadratic models (with or without provision for cell-killing) are fitted to the data on all cancers except leukaemia and on those five sites where a clear dose-response curve had previously been obtained (i.e., leukaemia, and cancers of the stomach, colon, lung, and female breast), a simple linear model fits the data on leukaemia, cancers of the stomach, lung and female breast, and all cancers except leukaemia better than the quadratic model and as well as the linear-quadratic model, as judged by the deviance ... Inclusion of cell-killing does not significantly improve the fit, except in one instance where leukaemia mortality under either the linear or linear-quadratic model fits somewhat better with a cell-killing term. These findings hold true both for organ absorbed doses and shielded kerma."
Nonetheless, on the very next page, UNSCEAR-88 appears to retreat from the linear model and to endorse use of the model which yields lower risk-estimates (Un88, p.480, Para.519): "Most current studies use a linear model for breast and thyroid cancer and a linear-quadratic model for other sites; these are the best-available models only, for the data do not really permit the validation of a specific model with confidence."
We note a real conflict between the UNSCEAR assertion and our own findings. We find that the data permit us to rule out the concave-upward dose-response with a great deal of confidence, and even to establish that the linear response is significantly inferior to the supra-linear dose-response.
Can it be that failure of UNSCEAR and its sources to use a "constant-cohort, dual-dosimetry" approach causes them to miss the supra-linearity? And what is the effect of their throwing out the 1950-1955 evidence? We will return to this after reviewing what BEIR-5 says about shape.
BEIR-5 on Shape :
Referring to the A-Bomb Study, BEIR-5 says in its Executive Summary (p.5): "The dose-dependent excess of mortality from all cancer other than leukemia, shows no departure from linearity in the range below 4 sievert (Sv), whereas the mortality data for leukemia are compatible with a linear-quadratic dose response relationship."
BEIR-5 restates this finding several times within the report, and shows no quadratic term in its best-fit equations for its several subdivisions of solid cancers (pp.168-170). Except for leukemia, all the BEIR-5 Cancer-Yields are based on the linear dose-response (Beir90, p.6).
Above, in our Part 2, we already discussed how BEIR-5 responded to its finding of supra-linearity: It threw out the high-dose groups. Our finding, however, is that supra-linearity is not limited to the high-dose groups. So we differ on shape with the BEIR Committee, too.
Can it be that failure of BEIR-5 to use a "constant-cohort, dual-dosimetry" approach causes it to miss the supra-linearity throughout the dose-range? And what is the effect upon shape of their throwing out the 1950-1955 evidence? And the effect upon shape of their throwing out cancers occurring after age 75?
First Crisis for the A-Bomb Study :
In first proposing that RERF support a "constant-cohort, dual-dosimetry" approach to the A-bomb database (Chapter 6), we predicted that problems of believability for new findings were certain to arise in the absence of such an approach.
We have shown above that the first crisis has arisen even earlier than we expected. A difference, or a change, in the shape of dose-response is a matter of the greatest seriousness. The entire A-Bomb Study's credibility can appropriately be called into question unless the reason for this difference can be fully traced and explained. And a comparison of apples with oranges permits no such explanation.
The simplistic suggestion can be made that all of the difference arises because three more years of follow-up (1983, 1984, 1985) have been provided for the abridged DS86 database used by the radiation committees (and by Muirhead; see Chapter 30) than for our "constant-cohort, dual-dosimetry" analysis for 1950-1982. It is possible that this is the explanation, and we wish to find out.
On matters of such importance, no one should have to speculate concerning what the truth is. If RERF provides the additional cancer-death data through 1985 for the T65DR cohorts used in this book, and reported on in RERF's TR-1-86 (91,231 persons), then we can calculate the up-dated dose-response for 1950-1985 in the T65DR and DS86 dosimetries, by our method of "constant-cohorts." If the results show that the last three years of follow-up have altered the shape of dose-response (after 12-years of a supra-linear shape), such a finding will be of profound importance.
But if the results show that dose-response remains supra-linear in both dosimetries with the "constant-cohort" method, despite the additional three years of follow-up, then a very serious problem exists. It would mean that the radiation committees' failure to find supra-linearity is caused by all the retroactive alterations, including deletion of 15,000 persons and shuffling of the remaining 75,000 into new cohorts.
A full explanation would be required of just how it could happen that a DS86 analysis by the "constant-cohort" method could possibly give a different answer from a DS86 analysis with the cohort-shuffling method currently in practice at RERF.
The credibility of the A-Bomb Study is going to hang in the balance until such questions are resolved. These questions and others which may arise in the future could all be resolved on a permanent basis if RERF itself would adopt a "constant-cohort, dual-dosimetry" approach along the lines we have proposed in Chapter 6.
-- Part 3, Our Finding :
Next, we are going to compare our own risk-estimates for acute-low and slow-low exposures with those of UNSCEAR-88 and BEIR-5. As usual, we will assemble the final values from the three sources afterwards.
Our own low-dose Lifetime Fatal Cancer-Yields come from Table 16-B for the A-Bomb Survivors and from Table 16-C for the United States population. These tables, unlike the tables in Chapters 13 and 14, are based on age-specific risk-coefficients -- as are the UNSCEAR-88 and BEIR-5 Cancer-Yields.
Our age-specific risk-coefficients or K-values for low-dose exposure incorporate (A) our finding from Chapter 14 that dose-response is supra-linear, with the highest cancer-risk per rad in the lowest dose-range, and (B) an interpolation within the short segment of the dose-range lying between 15 rads and zero rads.
Thus Equation (2) from our "Q.E.D." section applies, and there is no reason whatsoever to reduce our low-dose Cancer-Yields for slow delivery. They are proper risk-estimates for both acute-low and slow-low exposures.
LIFETIME FATAL CANCER-YIELD (excl. leuk.) for the dose-range 0-5 rads, any dose-rate. Number of fatal radiation-induced cancers per 10,000 persons, per rad.
Data are in next column.--------------- ------------------ Table 16-B, Table 16-C, A-bomb United States' survivors : population : T65DR = 31.65 T65DR = 26.64 DS86 = 30.43 DS86 = 25.56Values need doubling for exposure by X-rays.
-- Part 3, Radiation Committees :
Both UNSCEAR-88 and BEIR-5, having conceded that the best available human evidence shows a linear dose-response, nonetheless refuse to use the information. If they were to use it, their risk-estimates for acute-low and slow-low exposures would be exactly the same per rad as their moderate-to-high acute risk-estimates per rad.
Instead, both committees recommend dividing their risk-estimates by numbers like 2 up to 10 -- which is equivalent to reducing them by factors of 0.5 to 0.1. UNSCEAR-88 makes this recommendation for both acute-low and slow-low estimates, whereas BEIR-5 recommends the reduction only for slow-low exposures.
UNSCEAR 1988, Low-Dose Cancer-Yield :
UNSCEAR-88, in its executive summary, states (Un88, p.39, para.249): "In this Report, the problems in deriving risk coefficients at low doses and for low dose rates remain. The Committee agreed that there was a need for a reduction factor to modify the risks shown in Table 9 and Table 10 for low doses and low dose rates. [UNSCEAR's Tables 9 and 10 apply to 100 rads of organ absorbed dose delivered at high dose rate.] The Committee considered that such a factor certainly varies very widely with individual tumour type and with dose rate range. However, an appropriate range to be applied to total risk for low dose and low dose rate should lie between 2 and 10. The Committee intends to study this matter in detail in the near future." This same statement is repeated, almost verbatim, at page 494, para.623. In "typical situations," the number which UNSCEAR-88 recommends is five (Un88, p.491,para.602).
We cannot find any effort by UNSCEAR-88 to reconcile its endorsement of either "five" or "two to ten" with its statement, cited above, that "there is no clearly significant evidence of non-linearity" in the A-Bomb Study, so that "linear risk estimates are a reasonable summary of the dose-response."
Readers of our Chapter 22 will recognize the familiar "two to ten" range, suggested in 1980 by NCRP and based not on the human evidence already available then, but rather on other animals. UNSCEAR-88 (p.491, para.602) explicitly cites Ncrp80.
The UNSCEAR-88 multiplicative-model Cancer-Yield for acute exposure to 100 rads is 11.93 (see Part 2 of this chapter), excluding leukemia and based on a Japanese population. Thus, for acute-low and slow-low exposure, UNSCEAR's Cancer-Yields have the following range:
(11.93) / (2) = 5.96Values need doubling for exposure by X-rays.
(11.93) / (5) = 2.39
(11.93) / (10) = 1.19
BEIR-5, Low-Dose Cancer-Yield :
BEIR-5, referring to its Cancer-Yield for a single exposure to 10 rems, says in its Executive Summary (p.6): "For low LET radiation, accumulation of the same dose over weeks or months, however, is expected to reduce the lifetime risk appreciably, possibly by a factor of 2 or more."
In BEIR-5's Chapter 1 (p.22), the report acknowledges: "There are scant human data that allow an estimate of the dose-rate effectiveness factor (DREF)." The discussion procedes to confirm what we have explained in Chapters 22, 23, and earlier in this chapter -- that DREFS depend on dose-response having an upward-concave curvature. Then BEIR-5 mentions leukemia, the only malignancy where it is claiming a concave-upward dose-response. No comment is made that this one site-specific finding may be spurious, in view of BEIR's failure to find such curvature for any other subset of cancer with the same type of analysis. Instead, BEIR-5 reports (p.23) that the leukemia curvature indicates a DREF of 2.1.
After singling out leukemia from the human data, BEIR-5 completely retreats from human evidence -- evidence which shows that no reduction for slow dose-rate is appropriate. Instead, BEIR-5 focuses on laboratory animal studies. So its Table 1-4, "Summary of Dose-Rate Effectiveness Factors for Low-LET Radiation," consists of three types of entries: (1) DREF of 2.1 from human leukemia as analyzed by BEIR-5, (2) DREF of 2.0 to 2.5 from human leukemia as analyzed by BEIR-3 (a very questionable analysis; see our Chapter 22), and (3) the familiar DREF-range of 2-10 from laboratory animals. For the latter range, BEIR-5 lists "4" as the "single best estimate."
This evidence from other species is clearly the only basis for what BEIR-5 refers to as "the consensus" (Beir90, p28): "For low-LET radiations, the consensus is that decreasing the dose rate or dividing a given dose into a number of fractions spread over a period of time reduces the biological effectiveness." And indeed, as noted in Chapter 23 and above, we do not rule out the possibility that risk is reduced if a high total dose is delivered slowly. But BEIR-5 is recommending "dreffing" (reducing) the Cancer-Yield which applies to an acute dose of only ten rems, if the ten rems are accumulated slowly. We have shown, and BEIR-5 will admit at a later page, that there is no basis in either the human evidence or in logic for any such reduction.
Substitution of Non-Human for Human Data :
BEIR-5 addresses its retreat to the animal evidence in the most superficial way, in our opinion: "The committee felt strongly that its risk assessments should be based on human data to the extent that they were available and that animal data should be used only to address questions for which human data were unavailable or inadequate. Questions in the latter category included the RBE of neutrons and gamma rays and the effect of dose rate" (Beir90, p.55).
We call this explanation "superficial" because it offers no reason for regarding the A-Bomb Study as "inadequate" -- the very same study on which BEIR-5 relies for its leukemia DREF and most everything else in its risk-analysis (see Part 1 of this chapter, above).
And the statement by BEIR-5 certainly does not explain how one could possibly trust other species more than the human species, for evaluating the human dose-response. Everyone recognizes that "... the transfer of inferences from animal studies to humans is perilous" (Ncrp85, p.1), and that "Extrapolation from animal data to humans remains a difficult process of uncertain validity" (Ncrp85, p.36).
One cannot justify the substitution of non-human data for human data just because we have more of the non-human data. A mountain of stable and elegant animal data is worse than none at all, if it is mistakenly assumed to be relevant where it is not, and if it is used to replace the direct human evidence whose relevance is self-evident.
BEIR-5 offers no justification. The following quote shows that BEIR-5 makes a key recommendation which is admitted to be in conflict with the human evidence (Beir90, p.171):
"Since the risk models were derived primarily from data on acute exposures ... the application of these models to continuous low dose-rate exposures requires consideration of the dose rate effectiveness factor (DREF), as discussed in Chapter 1 [see quotations above]. For linear-quadratic models, there is an implicit dose-rate effect, since the quadratic contribution vanishes at low doses and, presumably, low dose-rates leaving only the linear term which is generally taken to reflect one-hit kinetics. The magnitude of this reduction is expressed by DREF values. For the leukemia data, a linear extrapolation indicates that the lifetime risks per unit bone marrow dose may be half as large for continuous low dose rate as for instantaneous high dose rate exposures. [We are splitting the BEIR-5 paragraph here.]
"For most other cancers in the LSS [A-Bomb Life Span Study], the quadratic contribution is nearly zero, and the estimated DREFS are near unity. Nevertheless, the committee judged that some account should be taken of dose rate effects and in Chapter 1 suggests a range of dose rate reduction factors that may be applicable."
The range suggested in BEIR's Chapter 1 (Table 1-4) is the familiar 1980 NCRP value of two to ten. At a later page (p.220), BEIR-5 mentions "2," whereas at page 6 it recommends "2 or more," and at page 23, it calls 2 to 4 the best estimates.
In the BEIR-5 statement above (p.171), which maintains the pattern documented in our Chapter 22, "nevertheless" introduces the key action. A scientific basis for this action escapes us.
The BEIR-5 Cancer-Yield for acute exposure to ten rads is 8.55 (see Part 2 of this chapter), excluding leukemia and based on a United States population. Thus, for slow-low exposure, BEIR-5's Cancer-Yields have the following range:
(8.55) / (2) = 4.28
(8.55) / (4) = 2.14
(8.55) / (10) = 0.86
BEIR-5 is explicit in suggesting that all its values may need doubling for X-ray exposure (Beir90, p.26, p.218, p.220).
-- Part 3, Discussion :
Now we will assemble all the values from above.
Gofman Estimate from A-Bomb Study:
LIFETIME FATAL CANCER-YIELD (excl. leuk.) for the dose-range 0-5 rads, any dose-rate. Number of fatal radiation-induced cancers per 10,000 persons, per rad.--------------- ------------------ Table 16-B, Table 16-C, A-bomb United States' survivors : population : T65DR = 31.65 T65DR = 26.64 DS86 = 30.43 DS86 = 25.56Values need doubling for exposure by X-rays.
The UNSCEAR-88 multiplicative-model Cancer-Yield for acute exposure to 100 rads is 11.93 (see Part 2 of this chapter), excluding leukemia and based on a Japanese population. Thus, after "dreffing" for acute-low and slow-low exposure, UNSCEAR's Cancer-Yields have the following range:
(11.93) / (2) = 5.96Values need doubling for exposure by X-rays.
(11.93) / (5) = 2.39
(11.93) / (10) = 1.19
The BEIR-5 Cancer-Yield for acute exposure to 10 rads is 8.55 (see Part 2 of this chapter), excluding leukemia and based on a United States population. Thus, after "dreffing" for slow-low exposure, BEIR-5's Cancer-Yields have the following range:
(8.55) / (2) = 4.28
(8.55) / (4) = 2.14
(8.55) / (10) = 0.86
BEIR-5 is explicit in suggesting that all its values may need doubling for X-ray exposure (Beir90, p.26, p.218, p.220).
Our analysis, above, indicates that the radiation committees are underestimating risk from slow-low exposure by up to 30-fold. It is easy to see how the close agreement in Part 2 converted into a big disparity. We began in Part 2 with a difference of about 1.5-fold. Our own estimate for low-dose exposure goes up by about a factor of two from its value at 45 rads, because of supra-linearity. So the disparity widens to 3-fold. And if the radiation committees cut their own estimates to one-tenth, in adjusting them from acute-moderate to slow-low exposure, then the disparity can widen to a factor as big as 30.
We have already discussed what is needed in order to resolve the difference over shape. As for our difference over "dreffing," there is no foreseeable reconciliation if one party is going to repudiate good human evidence and replace it by non-human evidence which gives a reduced risk-estimate.
It seems to me that BEIR-5 is trying to sit on both ends of a teeter-totter at the same time. At the one end, BEIR-5 takes the position that human dose-response is linear in the study on which it relies for most of its report. And at the other end of the see-saw, it repudiates its own statement of linearity by taking a dose rate reduction-factor.
BEIR-5 has produced neither evidence nor logic to support any speculation that the shape of human dose-response may suddenly change below ten rads. And BEIR-5 has correctly noted that if the shape were to change below 10 rads for reasons unknown, it could change in either the direction of increased risk or decreased risk per rad (Beir90, p.6, p.181). But the scientifically objective presumption, in the absence of contrary evidence or logic, is that the shape does not suddenly turn below 10 rads.
Thus we regard it not only as scientifically reasonable, but virtually obligatory, to use the reasonable presumption. And as far as we can tell, BEIR-5 is willing to use the linear model to estimate risk at acute doses below 10 rads, for it suggests the use of DREFS only in association with slow-low (not acute-low) exposures.
In the dose-range below ten rads, with respect to speculation about diminished risk if delivery is slow, we recommend close attention to track-analysis -- and particularly to "The Fallacy of Slow Delivery of Very Low Doses" in our Chapter 20.
For instance, our Table 20-M estimates that at a total tissue-dose from X-rays of one rad, cell-nuclei are experiencing only about one primary ionization track, on the average. Since, at the level of cell-nuclei, there is no conceivable dose-rate lower than one track at a time, it would approach absurdity to speculate that such a dose may be less hazardous if it is delivered slowly than if it is delivered all at once.4. Disproof of Any Safe Dose or Dose-Rate
As demonstrated in our Chapters 24 and 34, influential segments of the radiation community have been speculating in favor of a "threshold" -- the notion that low doses and dose-rates may be completely safe ("without effect"). The U.S. Department of Energy calls this the "zero risk" model (Doe87, p.J.8). It received widespread attention when an abbreviated version of Doe87 was published by the journal Science (Ansp88).
Indeed, opportunities for disseminating such speculations have increased, because slow-low exposure of populations has been much in the news after the Chernobyl accident and after the series of recent revelations about radioactive pollution inside, outside, and beneath many Department of Energy facilities. The estimated cost to clean up DOE's contamination ranges from $63 billion (DOE estimate) to $175 billion (Government Accounting Office estimate), according to the press (Nyt88; Nyt89a).
The BEIR-3 Report of 1980 (see our Chapter 34) "chose not to include the class of functions with a threshold, i.e., functions in which the cancer risk is zero up to some positive value on the dose-scale" (Beir80, p.181).
At the same time, the BEIR-3 Committee encouraged threshold speculation (A) by stating "It is by no means clear whether dose rates of gamma or x radiation of about 100 mrads/yr are in any way detrimental to exposed people" (Beir80, p.139), and by stating that the BEIR-3 Committee itself felt "uncertainty as to whether a total dose of, say, 1 rad would have any effect at all" (Beir80, p.193), and (B) by stating that it was not possible to settle the question within the available evidence (Beir80, p.22).
We disagreed (Go81). And we disagree today even more emphatically.
-- Part 4, Our Finding :
In Chapters 18 through 21 of this book, plus supporting Chapters 32 and 33, we prove beyond reasonable doubt and by any reasonable standard that no safe dose or dose-rate exists for the human with respect to radiogenic cancer. Our disproof of a threshold is based on the human evidence (much of which existed in 1980, too).
Our finding means that when very large populations are exposed to very small increments of ionizing radiation, the cancer-fatalities can be enormous even though each individual's personal risk is low. Chapter 24 uses the small individual doses from the Chernobyl explosion as just an illustration. Chapter 3 (page 1) mentions another illustration -- our 1985 estimate of the cancer-deaths induced in the United States by unnecessarily high doses given during diagnostic X-ray exams.
Disproof of any safe dose or dose-rate means that such cancer-deaths are not "hypothetical" and not "imaginary." They are real.
Thus, there is no issue of greater importance to the public and to the radiation community, than the threshold issue.
-- Part 4, Radiation Committees :
UNSCEAR 1988 on Threshold :
On the threshold issue, the UNSCEAR-88 Report (p.411, para.26) refers its readers to its previous report of 1986, Annex B. And there one finds various statements like the ones quoted in our Chapter 34.
UNSCEAR-86, p.166, para.3: "Although the absence of the threshold is often assumed, this has not been proved for any form of radiation-induced malignancy and must be regarded as a working hypothesis ... Proving or disproving a threshold below the levels of direct observation may be impossible, due to statistical fluctuations of the spontaneous level and of the presumably induced response ..."
At a later page, UNSCEAR-86 makes a statement (about dose-response curvature) which suggests that UNSCEAR should approve of our own approach to the threshold question: "What may happen at the low doses, where direct information is lacking, may only be inferred from a combination of empirical data and theoretical assumptions, linked together into some models of radiation action" (Un86, p.241, para.474).
Our disproof of any safe dose or dose-rate takes the the human epidemiological observations at low tissue-doses, and combines them with track-analysis (number of primary ionization tracks per cell-nucleus), to show that the empirical observations do tell us what is happening at even lower tissue-doses.
BEIR-5 on Threshold :
The repair of carcinogenic lesions, inflicted by a single primary ionization track upon genetic molecules in a cell-nucleus, is a key issue in speculations about a safe dose or dose-rate of ionizing radiation.
The BEIR-5 Committee alludes to the issue very early in its report (Beir90, p.4):
"Of the various types of biomedical effects that may result from irradiation at low doses and low dose rates, alterations of genes and chromosomes remain the best documented. Recent studies of these alterations in cells of various types, including human lymphocytes, have extended our knowledge of the relevant mechanisms and dose-response relationships. In spite of evidence that the molecular lesions which give rise to somatic and genetic damage can be repaired to a considerable degree, the new data do not contradict the hypothesis, at least with respect to cancer induction and hereditary genetic effects, that the frequency of such effects increases with low-level radiation as a linear, nonthreshold function of the dose."
However, the statement which is likely to be quoted by several segments of the radiation community is the statement which the BEIR-5 Committee provides much later (p.181):
"... epidemiologic data cannot rigorously exclude the existence of a threshold in the millisievert dose range. Thus the possibility that there may be no risks from exposures comparable to external natural background radiation cannot be ruled out. At such low doses and dose rates, it must be acknowledged that the lower limit of the range of uncertainty in the risk estimates extends to zero."
We are disappointed that BEIR-5 would issue such a statement without first testing the epidemiologic evidence against a track-analysis of the type demonstrated in this book.
-- Part 4, Discussion :
Readers of Chapters 24 and 34 will not be surprised if there is great resistance in parts of the radiation community to our disproof of any safe dose or dose-rate.
If it is claimed by anyone that the nine human epidemiological studies in our disproof do not constitute enough evidence, such a claim would establish quite a contrast of standards, for no human evidence is demanded in order for the radiation committees to recommend reducing risk-estimates (by "dreffing," as discussed in Part 3). Indeed, the recommendation is made contrary to good, human evidence.
For all the reasons stated in our Chapter 21, we think that we have disproven any safe dose or dose-rate, beyond reasonable doubt. But we do not expect readers to accept anyone's assertion on faith alone. The entire case has been laid out -- from the evidence, step-by-step, to the conclusion -- for readers to judge for themselves.5. Some Practical Implications for Human Health
How seriously need we take ionizing radiation, as a human carcinogen?
There is just no doubt that ionizing radiation is a cause of some unrepaired injuries to our genetic material (DNA and chromosomes). There is a vast body of work connecting genetic anomalies with human cancer. Indeed, the Nobel Prize in Medicine has just been awarded for work showing that disruption of our genes can trigger cancer. And the human epidemiological reality-check leaves no doubt that exposure of people to ionizing radiation, even at the lowest possible doses and dose-rates, results in excess fatal cancer.
The Most Important Single Carcinogen ?
A prominent member of the radiation community, Rosalyn Yalow (see Chapter 34), has asserted that "... exposure to ionizing radiation is only weakly carcinogenic" (Ya88, p.11). Weak on the basis of whose estimate? The question really matters, when estimates differ by factors of 2, 5, 10, 15, 20, 25, or even 30.
And weak compared with what?
The separate contributions from other carcinogens to the population's total cancer mortality are hardly quantified at all. My estimates of the risk-per-rad from ionizing radiation are consistent with some 15 to 25 percent of all human cancer being caused by ionizing radiation (see box). Radon injects considerable uncertainty into the range. Ionizing radiation may even turn out to be the most important single carcinogen to which large numbers of humans are actually exposed. No one can possibly be sure yet, in the absence of equally good epidemiological data on all the other human carcinogens and on the magnitude of human exposure to them.
"A Ball-Park Estimate" : For U.S. Population
Col.A Col.B Col.C Col. D Col.E Source Dose(rems) Dose(rems) Lifetime Fatal Fatal Cancers / (Annual) (60 year,cumulative) Cancer-Yield 10,000 persons Radon 0.2 12.0 26 312 Other natural 0.094 5.64 26 147 Medical X-rays 0.039 2.34 26 61 All other 0.024 1.44 26 37 TOTAL EXPECTED CANCERS PRODUCED IN 10,000 PERSONS = 557 Radiation-induced share of cancer-mortality = (557/2200)(100%) = 25.3% .
Other Human Carcinogens :
Even if human carcinogens could be reliably ranked, we certainly do not wish to imply that only the Number One killer deserves to be taken seriously. Far from it.
And we are troubled by the implications of what we see, in the radiation field, for the possible chemical carcinogens. Is the position going to be taken that it is acceptable to release "low-levels" of possible chemical carcinogens into the common indoor and outdoor environment, unless the public can prove from direct human epidemiological evidence for each of them that there is no safe dose?
Certain unique features of ionizing radiation permit us to know more about it than about other human carcinogens. For instance, in this field of research, we do not need to depend on possibly irrelevant, and therefore eternally inconclusive, evidence from non-human species or cell-studies. Without conducting immoral human experimentation, human data exist for ionizing radiation because it is widely used in medicine, diagnostically and therapeutically. In addition, as a result of the two atomic bombings in Japan, genuinely comparable groups of humans exist who were exposed to very different dose-levels. This situation is important in many ways (discovering the shape of dose-response, proving causality beyond a reasonable doubt, etc.), and yet the situation is unlikely to occur for many other carcinogens. And lastly, the unique physical properties of ionizing radiation make it possible to prove that there is no safe dose or dose-rate, even in the absence of human studies conducted at the lowest conceivable doses.
A Giant Gamble with Human Health ?
When we think over the fierce resistance to giving up the threshold idea with respect to ionizing radiation, even when there is such compelling human evidence against any safe dose or dose-rate, we wonder what behavior will prevail with respect to all those chemical substances where compelling human epidemiological evidence is lacking.
In Chapter 24, with respect to ionizing radiation, we have mentioned that proposals are pending to omit very low-dose exposure entirely from evaluations of human health effects ("de minimis non curat lex," or "the law does not concern itself with trifles"), and to treat a large share of "low-level" radioactive wastes (possibly some consumer items too) as if they were not radioactive at all -- "below regulatory concern."
If such proposals prevail, it is self-evident that less of the past radioactive contamination will be cleaned up, that future nuclear pollution will increase, and that human exposures to ionizing radiation will rise.
The "Global Perspective" :
The UNSCEAR 1988 Committee has raised the issue of public perception of radiation risk -- a timely issue after the Chernobyl nuclear accident has called worldwide attention to such risk. UNSCEAR makes no estimate itself of the cancer-fatalities which will occur from the accident, but it comments (Un88, p.42, para.268):
"... the way in which basic scientific facts are presented influences the impression they give. For example, thousands of cancer deaths from a single accident would undoubtedly be a high number of deaths. However, since such deaths could be expected to occur over a long period of time, the annual incidence will be low. This means a very small increase of the normal incidence of cancer, an increase which is not expected to be noticeable in health statistics. This shows that it is possible, by selecting the form of presentation, to convey different impressions."
Not surprisingly, parts of the radiation community have been extremely active in trying to shape public perception of the Chernobyl accident. In Chapter 24, we documented the repeated comparison of any estimate of people receiving a cancer-death from Chernobyl, with the far greater number of people who will die of cancer anyway in the northern hemisphere during the next 50 to 100 years.
The authors of the Department of Energy's Chernobyl assessment call this the "global perspective" (Ansp88, p.1513). If there is another accident -- or equivalent nuclear pollution from intentional releases -- we expect to hear about the "inter-stellar perspective."
To illustrate the consequences of "giving" small per-capita doses to entire populations, we presented some realistic estimates of Chernobyl's all-time cancer-consequences.
The range in Table 24-B, based on Cancer-Yields from both Gofman and from RERF analysts, is from 140,000 to 475,000 fatal cancers (leukemia excluded). Because such estimates come from very low per-capita doses received by hundreds of millions of people, such estimates could just be erased from the slate of a benefit-risk analysis, if risk-evaluation were ever to exclude all slow-low exposures (as in "de minimis" proposals).
Just One Part in a Thousand ?
It may sound like a trifle to put only one part per thousand of a poison into the environment, but we will show what one part per thousand means with respect to radioactive cesium.
The cesium-137 produced each year by a 1000-megawatt (electrical) nuclear power plant amounts to nearly 4 million curies. Since its radioactive half-life is 30.2 years, very little of it decays during a year.
The Chernobyl reactor contained a two-year cesium-inventory of about 8 million curies. Recent estimates are that the Chernobyl reactor released about 2.5 million curies of cesium-137, which is equivalent to (2.5 / 4.0) or 62.5 % of a one-year inventory.
Now let us consider 100 large nuclear power plants each operating in the USA for a lifespan of about 25 years each. Call "A" the yearly cesium-137 production by one plant. Then 100A = the yearly production by 100 plants. Lifetime production = 25 yrs x 100A/year = 2,500A. 99.9 % containment = release of 1 part per 1,000. With 99.9 % perfect containment, loss = 2.5A. Chernobyl lost 0.625A. The ratio of 2.5A and 0.625A is 4.0.
This ratio, 4, has an enormous meaning. It means that achieving 99.9 % perfect containment of the cesium-137 produced by 100 plants during 25 years of operation, through all steps of the cesium's handling up through final burial, would still result in cesium-137 contamination equivalent in curies to 4 Chernobyl accidents.
Worldwide, there are about 400 plants underway, so the same scenario (99.9 % perfection in containing cesium) would mean cesium-loss equivalent to 16 Chernobyl accidents per 25 years of operation. And this assault on human health could occur without blowing the roof off any single plant.
Best Estimates ... Semi-Prudence :
The stakes in the correct evaluation of cancer-risk from low-dose exposure extend far beyond one spectacular accident like Chernobyl. Not only do such evaluations affect hundreds of millions of medical and dental patients, and millions of occupationally exposed workers, but correct evaluation necessarily affects the decisions which will determine the ultimate and aggregate levels of radioactive pollution, everywhere, from current and contemplated nuclear activities worldwide.
It is possible that new evidence developing in the future will show that our estimates in this book, of cancer-risk from low-dose, low-LET ionizing radiation, are too high -- and it is equally possible that new evidence will show that our estimates are too low. In other words, there is as much chance that sampling variation and forecasting are producing underestimates of hazard as overestimates.
Pending future evidence, it is scientifically appropriate to produce and disseminate the best risk-estimates which come from the available human epidemiological evidence of good quality.
But we will repeat a warning.
What is scientifically appropriate behavior is only semi-prudent with regard to public health protection. True prudence with respect to human health would require the operating assumption that current uncertainties in sampling and forecasting are causing us to underestimate the real risk.