REDUCED RADIO-BIOLOGICAL EFFECTIVENESS AT LOW-RATE, LOW-DOSE EXPOSURES [DREF]
: An Unwarranted Conjecture
Rudi H. Nussbaum, Portland State University, Portland
OR 97205-0751/USA
and Wolfgang Köhnlein, Universität Münster/Germany
Historical Perspective
After the use of the A-bombs and the follow-up of its after-effects among a
large group of survivors of Hiroshima and Nagasaki, (the LSS cohort), there has
never been any doubt about the disastrous short- and long-term effects of
exposures to high doses of radiation, nor about the numerical relation between
high-dose exposure and excess risk. However, subsequent public acceptance of
escalating billion-dollar investment in nuclear weapons production, in civilian
nuclear energy and also in nuclear medicine, was predicated on the confident
assurances of enthusiastic radiation experts that added exposures at dose levels
acceptable to industry, would not be found detrimental to human health, they
might even be beneficial !
Premature evaluations of the delayed effects of radiation among the Japanese
survivors, as well as transference to humans of radiation effects in animals,
lead to strongly-held optimistic convictions. For decades, any challenge to
these tenets based on extrapolations by epidemiologic studies among populations
with occupational and environmental exposures were received with enmity and
rejection rather than as opportunities to test new hypotheses and increase our
understanding. Nevertheless, optimistic expectations had to be scaled back
continuously as a result of long-delayed excess cancers among the Japanese
survivor population.
Introduction
Direct Information on the relation between human cancer induction and radiation
dose, in particular at low doses and low dose-rates, should be considered as the
most reliable foundation for estimating occupational and environmental exposure
risks from radiation releases during the various stages in nuclear technology -
from uranium mining, to weapons testing, waste disposal and decommissioning.
During these various phases, which include accidents, workers and the general
public can be at risk by external and/or internal exposure to various ionizing
radiations. Historically, at the time of most rapid build-up of this technology,
when exposure standards for radiation protection had to be formulated, there
were several major barriers to such a direct approach. First, there has been a
scarcity of epidemiological studies among occupationally or environmentally
exposed populations with sufficiently long follow-up times. Decades of follow-up
are required to allow for long and varying periods of latencies for most
malignancies. In the context of the times authoritative scientists asserted that
low-dose studies would never be able to detect significant excess radiogenic
incidence or mortality.
Probably the most important barrier to direct information on human cancer
induction and radiation dose was the concentration in radiation epidemiology on
the post 1950 follow-up of a large population of Japanese A Bomb Survivors.
Early findings of the A-bomb study, strongly weighted toward medium to high
doses, in combination with high-dose animal data, have continued to dominate the
thinking about radiation health effects. Over recent decades, in view of
long-delayed and still increasing excess cancer mortalities among the LSS cohort
of A-bomb survivors, official cancer risks per unit exposure have had to be
revised upward several times. Significantly, the models postulated for
extrapolating A-bomb exposure risks downward to the occupationally relevant dose
range, are based on transference of observed dose- and dose-rate effects in
animals or animal cells under exposures to very high doses (several hundreds of
cGy). A majority of radiation experts used these extrapolation models from
animal and radio-biological findings to crystallize into firm expectations of
much reduced biological effects per unit dose at low doses and low dose-rates
(dose fractionation) in people.
A consequence of this mind set was that any epidemiologic evidence from studies
of populations with occupational and environmental exposures most relevant to
risk assessment, which were inconsistent with the above tenets, have routinely
been received with angry rejection by a majority of experts and established
radiation effects commissions. These groups refused to recognize that apparent
discrepancies in findings among vastly different populations could serve as
opportunities to test alternate hypotheses and to gain insight into poorly
understood physico-chemical interactions of ionizing radiation with human
health. Focus on the paradigm has meant that much less is known by radiation
experts about the potentially much wider spectrum of health effects related to
distributed or selectively concentrated radio-isotopes ingested into the body.
In part, this is due to the lack of reliable methods of internal dosimetry. Yet
many ailments are now showing up among atomic veterans, so-called downwinders of
weapons production plants, and populations exposed to the Chernobyl fallout.
Nevertheless, official scientific bodies have repeatedly denied any possible
connection with radiation.
Thus, two basic assumptions have endured in slightly varying versions in all
recent authoritative radiation effects reports such as UNSCEAR (1988), BEIR V
(1990) or ICRP (1990) :
(1) that a linear, no-threshold dose-effect relation, fitted to the A-bomb
mortality data over the complete dose range 0 - 400 cGy (Fig. 1, curve L) would
overestimate occupational and environmental risks per unit dose by at least a
factor two or more. Hence, low-dose risk (equivalent to the slope of the curve)
should rather be determined from a concave dose-effect relation with steadily
reduced slope at reduced dose, approximating the existence of a threshold dose
range (Fig. 1, curve Q).
(2) that independent of the relation describing low dose effects, the
"biological effectiveness" per unit dose is considerably lower (by at
least a factor two) for fractionated or extended than for single exposures.
Therefore, reductions of linear risk estimates by so-called "Dose and Dose
Rate Effectiveness Factors" or DREFs (ICRP),[4] (UNSCEAR) [2] between 2-10
have remained part of risk evaluations by most official radiation protection
commissions. For example, the authoritative BEIR V report (p. 22) states:
"There are scant human data that allow an estimate of the dose-rate
effectiveness factor." Nevertheless, in its Executive Summary, the most
quoted part of this report, the committee recommends use of a DREF of at least
two for environmental and occupational exposures [3]. Again, the 1990 ICRP
Recomendations [4] incorporate a DREF of two in their tables of occupational
risk estimates. The primary sources for the above two tenets (1) and (2) had
been the concave dose relation for leukemia mortalities found at very high doses
among the A-bomb survivors and animal, as well as animal cell studies at varying
dose rates and total doses of several hundred cGy.
As an alternative to the representation of dose-effect relations as shown in
Fig. 1, a more effective distinction between different risk models can be
obtained by a loglog plot of excess relative risk per unit dose DR/DD (essentialiy
the slope of the dose response curve) versus dose for different
dose-dependencies (Fig. 2). Let us now examine the above listed persistent
assumptions (1) and (2) in the light of some relevant recent research outcomes:
A-bomb Survivors (LSS cohort)
1. A linear model of risk (L-model, Fig. 1, curve L or Fig. 2, curve 1) gives an
excellent fit to the A-bomb data, when they are restricted to doses <200 cGy,
both for leukemia and for all cancers. Adding a positive quadratic term (LQ-model,
see Fig. 1, curve L or Fig. 2, curve 2 does not improve the fit. In fact, any
DREF factor of much above one to correct for a presumed so-called "linear
extrapolation overestimation" of risk, is quite inconsistent with the data,
according to RERF scientists [5,6].
2. Kerma doses below 5 cGy and probably as low as 1.6 cGy have produced excess
cases of acute lymphoid and chronic myeloid leukemia among A-bomb survivors [7].
Certainly, such a significant risk at very low single doses of exposure leaves
little room for a "safe" (very low slope) range in the dose-effect
relation.
3. When Aggregate cancer mortalities 1950-85 are plotted against mean kerma or
organ absorbed dose for the dose subcohorts in the 0 - 100 cGy range, we find a
strong indication for an initialiy much steeper slope below 10 cGy in the
dose-effect relation (Figs. 3, 4), exactly the opposite of the postulated
concave curve, illustrating the generally accepted LQ-model [8, 9]. Given the
confidence limits of the 1950 - 85 cancer mortality data for the lowest LSS dose
groups, together with documented nonuniform under-estimates of the mean doses
for the control and the lowest exposure groups, a linear model with a single
slope down to zero dose can not be excluded on the 95% level of confidence
(p=0.05). However, consistency of our linear slope for the range 6- 69 cGy with
that obtained by RERF scientists for the entire dose range under 200 cGy,
strongly suggests a steeper initial slope, thus higher risk per unit dose, for
the very low dose range 0 - 10 cGy resulting from a single A-bomb exposure (see
Fig. 1, curve P; Fig. 2, curve 3 and Fig. 5) [8,9].
4. Another independent analysis of the same low-dose A-bomb data suggests a best
fit to a convex mortality-dose relation of the form m = mo + aDb (b
5. One of the chief analysts of RERF data, D.A. Pierce, applied RERF's model,
stratified for background rates according to age at time of bombing, sex, city
and follow-up period, to the LSS dose groups below 100 cGy. He confirmed
improvement in the fit (p=0.25) for that dose range, if instead of simple
proportionality with dose, he assumed excess relative risk to be proportional to
the square root of dose (Fig. 5) [11]. The same convex dose dependence had
earlier been suggested for the excess radiation risk in a mortality analysis of
Hanford nuclear workers at very low accumulated doses (see below and Fig. 5)
[12]. On the basis of stringent statistical criteria, such convex (supra-Iinear)
dose-effect curves from studies of these two very different populations are
strongly suggestive, but not compelling. However, together with additional
epidemiologic and radio-mutagenic evidence, a picture emerges which by its
consistency, supports the hypothesis of an increased biological effectiveness
per unit dose at very low doses, contrary to assumption (1) above.
Other Populations
1. There is now a large body of studies on childhood cancers following prenatal
exposures, the first of which was Stewart and Kneale's Oxford Survey of
Childhood Cancers. They all show significant excess risks for X-ray exposures
down to fractions of a cGy,[13,14,15,16] thus leaving no room for a
"safe" dose range for human exposure.
2. A 28-50 year follow-up study of over 31,000 Canadian women, following X-ray
fluoroscopy as part of treatment for Tuberculosis, with doses ranging from 10
cGy or less per treatment to over 1,000 cGy accumulated dose, showed breast
cancer mortalities most consistent with a simple linear dose-response model over
the entire dose range [17]. Considering the fractionation of dose in that study
and the large range of accumulated doses, there is no support for either the
quasi-threshold notion of the LQ model at low doses, or any reduced cancer
induction at fractionated exposures in this study of female breast tissue.
3. The occupational radiation study by Mancuso, Stewart and Kneale (MSK) of
Hanford nuclear workers, already mentioned above [12], was rejected and maligned
by the community of radiation experts when it suggested that official risk
estimates, based on extrapolations from the A-bomb study, implicitly using the
above tenets, were off by at least a factor ten. Recent reanalysis by the same
researchers, including updated Hanford mortality statistics, reconfirmed most of
the earlier findings. A just completed DOE-sponsored mortality analysis for Oak
Ridge nuclear workers 1943 - 84, using more conventional statistical methods,
confirmed the much higher radiation risks at very low and protracted exposures
found in the MSK Hanford study [18]. These US worker data are also consistent
with some studies of British nuclear workers [19] (Table 1 and Fig. 5). The most
recent British nuclear worker study, involving a much larger study population,
while broadly consistent with the US studies, would need further statistical
refinement to be comparable in detail to the Oak Ridge study [18]. However, that
British study, too, found evidence for an association between radiation exposure
and moriality from cancer, in particular leukemia and multiple myeloma, at very
low accumulated doses, protracted over long periods of time [20]. Neither of
these recent epidemiological data do allow a statisticaily robust distinction
between a linear and a convex dose-response relation. However, the values of
excess relative risk DR/DD versus mean dose D in the representation suggested in
Figs. 2 and 5 for three of the nuclear worker studies mentioned, as well as for
the low-dose A-bomb data (see also Figs. 3, 4) are consistent with a convex
(supra-linear) dose-effect relation (Fig. 5). Leaving the question of the
correct shape of this curve aside for the moment (it is likely different for
different populations), the aggregate of findings does establish significant
excess risks for any dose over and above background and at background exposure
rates, even in populations selected for good health (in US studies a 25% lower
mortallty for all causes among nuclear workers compared to the general
population) which clearly contradicts both tenets (1) and (2) and thus
undermines the concept of "safe" occupational or environmental
exposure levels. In addition, the mutually consistent magnitudes of the excess
risk per cGy in these worker studies, in , comparison with the somewhat lower
risks from the low-dose A-bomb data for a single exposure above 6 cGy mean dose
(Table 1 and Fig, 5), suggest an increased, not a reduced risk per unit dose at
very low and fractionated exposures. All serious challenges to the Oak Ridge
data, attempting to dismiss them as "singular" or seeking to
"reconcile" them with the A-bomb data,. have been effectively refuted
by the authors of the study [18].
4. Also contrary to assumed reduced biological effectiveness at lower
dose-rates, a survey of lung cancer studies among uranium miners found that
relative risk for internal alpha radiation exposures increases significantly
with decreasing exposure rates [21].
Mutational and Animal Data
While epidemiological studies involve whole organisms in the context of
complicated co-exposures, selection, susceptibility factors, etc., studies of
mutagenesis deal with greatly simplified systems, usually cell cuItures, Thus,
in these studies there is greater opportunity for experlmental control and
reliable measurement, however, their outcomes yield only an indirect and poorly
known link to factors determining human health. Some animal studies are
considered to provide closer models for human reactions to radiation exposures.
High-dose animal and radio-biological data on cell mutations have been the
primary evidence cited for the two-fold DREF hypotheses (1) and (2). Several
recent findings in these fields, however, contradict these assumptions:
1. Waldren et al. [22] have demonstrated that conventional methods for measuring
mutagenesis in mammallan cells seriously underestimate the contribution of
radiation to cancer and genetic diseases. They observed at least a 200-fold
higher mutation frequency in the 0-50 cGy range than some previous conventional
studies. Also, they find a supra-Iinear mutation-rate-dose relation (Fig.6).
This study deserves confirmation, in particular at very low doses and varying
dose-rates.
2. Grosovsky and Little studied mutagenesis in human lymphoblast cells over a
wide range of doses and dose-rates. They found a linear dose dependence and no
indication of a change in mutation frequency per unit dose between single and
fractionated doses (Fig. 7) [23, 24]. Also, exposures to tritiated water and
X-rays over a range of dose rates yielded no deviation from a proportional
relation between mutation frequency and total dose [25].
3. Little showed significant differences, with regard to inhibitory effects of
DMSO, dose response and the effects of changes in dose-rate for radiogenic
induction of mutations in rodent, compared to human cells. While he confirms
reduced effects in rodent cells with reduced dose rates at high total doses, no
such effects are found in human cells and the dose response is linear in the
range 0 - 250 cGy (Fig. 8a, b) [24]. These results make the transference of
rodent exposure findings to humans extremely questionable.
4. A recent radiation study on 378 beagle dogs with accumulated doses between
450 - 3000 cGy at varying rates of 3.8 - 26.3 cGy/day shows no relationship
between tumor mortality and dose rate but a clear linear relation with
accumulated dose [26].
Conclusion
Vigorous rejection by a majority of radiation experts of suggestions that the
two long-held tenets mentioned above are inconsistent with current human data
epidemiologic or mutagenic - may in part be based on the recognition that a
convex dose-effect relation, especially at very low doses, cannot be reconciled
with the long-known linear mutagenic effect of ionizing radiation, down to the
lowest doses. In fact, any observed increased biological effectiveness per unit
dose at near-background exposures which may well rapidly saturate at doses below
10 cGy, followed by a linear relation (as suggested by the A-bomb mortalities),
would indicate the existence of one or several hitherto unrecognized pathways
toward cancer induction, especially effective at very low exposures (where the
linear effect is relatively small) and highly non-linear in dose [27] *'
(footnote)
This is not the only evidence that our understanding of the radiation -
chromosome interactions is rather incomplete. Other examples are: the
association between the father's exposure to relatively low doses of external
radiation and leukemia in his offspring [28], or genetic abnormalities that
express themseives only after several generations of cell divisions in mouse
cells after exposure to alpha particles [29] and also in hamster cells after
exposure to X-rays [30, 31]. Finally, a new mechanism of inflammatory reactions
in human blood has been found in the ultra-low dose range 5.4 - 235 microGy,
with a linear dose-effect relation up to 100 microGy, followed by a plateau.
This mechanism in blood is quite distinct from the effects of exposures at
higher doses [32].
To summarize: for decades radiation researchers have focused their
attention somewhat myopically on delayed cancer and genetic effects among A-bomb
survivors - a highly selected group of individuals - and on much animal and
radio-biological research at very high doses. Specific low-dose or low dose-rate
studies of populations or human ceils, that suggested that ionizing radiation
might actually have an increased rather than a reduced biological effectiveness
at low doses, have either been rejected outright, or ignored in reviews of
radiation health effects. Only very recently have there been persistent
suggestions in the literature that such unexpected effects which clash with
firmly held beliefs among a majority of radiation experts, might involve
hitherto unknown and rapidly saturating complex radio-biophysical or
radiobiochemical mechanisms at very low doses, very different from the
well-known mutational effects, proportional to dose,
" Two recent independent studies of correiations between neo-natal deaths
and radioactive contamination of the environment by nuclear tests and the
Chernobyl explosion, are consistent with the implications of Petkau's work on
the effect on cell membranes of low-dose, low-rate exposures. lt is wellknown
that membrane functions play a crucial role in the effective operation of the
human immune system. All of these studies strongljy suggest much higher low-dose
effects than a linear extrapolation from higher doses would predict. (see also
refs. 9, 10, 12, 18,19-21): LUNING G, SCHEER J, SCHMIDT M and ZIGGEL H. Early
infant mortality in West Germany before and alter Chertiobyl. The Lancet, Nov.4,
1989: 1081-1083. WHYTE RK. First day neonatal mortality since 1935:
re-examination of the Cross hypothesis. British Medical Journal: 304, Febr.
8,1992. see also the book by GOULD JM and GOLDMAN BA, Deadly Deceit,- Low Level
Radiation, High Level Cover-up. New York, Four Walls Eight Windows, 1990.
presented at First International Conference of the Society for Radiation
Protection (Germany), Kiel, Febr. 28 -March 1, 1992 published 1993.
References
[1]. LAND CE.Estimating cancer risks from low
doses of ionizing radiation. Science 209: 1197-1202 (1980).
[2]. UNITED NATIONS COMMITTEE ON THE EFFECTS OF IONIZING RADIATION. Sources,
Effects and Risks of lonizing Radiation. New York, NY: United Nations, 1988.
[3]. BEIR V: Health Effects of Exposure to Low Levels of lonizing Radiations,
Board on Radiation Effects Research, Commission on Life Sciences, National
Research Council. Washington, DC: National Academy Press, 1990.
[4]. INTFRNATIONAL COMMISSION ON RADIOLOGICAL PROTECTION. Recom-mendations of
the ICRP. Publication 60, Pergamon Press, London, 1991.
[5]. SHIMIZU Y, KATO H, SCHULL WJ. Life Span Study Report 11. Part 2. Cancer
mortality in the years 1950-1985 based on the recently revised doses (DS86).
RERF TR 5-88. Radiation Effects Research Foundation, Hiroshima, 1988 and Radiat.
Res. 126: 36-42 (1991).
[6]. PIERCE DA, VAETH M. Cancer risk estimation from the A-bomb survivors:
Extrapolation to low doses, use of relative risk models, and other
uncertainties. RERF CR 2-89. Rad. Eff. Res. Found., Hiroshima, 1989, ---- The
shape of the cancer mortality dose-response curve for atomic survivors. RERF TR
7-89. Rad. Eff. Res. Found., Hiroshima, 1989 and Radiat. Res. 121: 1 20-41
(1990).
[7]. TOMONAGA M, MATSUO T, CARTER RL. A-bomb Irradiation and leukernia types: An
update. RERF (Update, 3(4), winter 1991/92. Rad. Eff. Res. Foundation,
Hiroshima, 1992.
[8]. NUSSBAUM RH, KOHNLFIN W, BELSFY RF. Die neueste Krebsstatistik der
Hiroshima-Nagasaki Überlebenden. Med Klin 86:99-108 (1 991)
[9]. KOHNLEIN W , NUSSBAUM RH. Reassessment of radiogenic cancer risk and
mutagenesis at low doses of ionizing radiation. Adv Mutagen Res 3:53-80 (1991).
[I0]. GOFMAN JW. Radiation-induced cancer from low-dose exposure: An independent
analysis. Committee for Nuclear Responsibility Book Division, San Francisco,
1990.
[11]. PIERCE DA, private communication, 1991.
[12]. KNEALE GW, MANCUSO TF, STEWART AM. Hanford Radiation Study III: A cohort
study of the cancer risks from radiation to workers at Hanford (1944-77 Deaths)
by the method of regression models in life-tables. Brit J Ind Med 38 :156-1 66
(1981); and unpublished update, 1991.
[13]. KNEALE GW, STEWART AM. Pre-natal X-rays and cancers: Further tests of OSCC
data. Health Phys 51:369-376 (1986); 53:200 (1987)
[15]. GILMAN FA, KNEALE GW, KNOX EG, STEWART AM. Pregnancy X-rays and childhood
cancers: Effercts of exposure age and radiation dose. J Radioi Prot (GB) (1988).
[16]. BITHELL JF, STILLER CA . A new calculation of the radiogenic risk of
obstetric x-raying. Stat Medicine 7:857-864 (1988)
[17]. MILLER AB, HOWE GR, SHERMAN GJ, LINDSAY JP, YAFFE MY, DINNER PJ, RISCH HA,
PRESTON DL. Mortality from breast cancer after irradiation during fluoroscopic
examinations in patients being treated for tuberculosis. N Engl J Med
321:1285-89 (1989). See also discussion in ref. (10), p. 22-14, 15,
[18]. WING S, SHY CM, WOOD JL, WOLF S, CAGL DL, FROME EL. Mortality among
workers at Oak Ridge National Laboratory: Evidence of radiation effects in
follow-up through 1984. JAMA 266:1397-1402 (1 991). Discussion: JAMA 266:653-654
(1991): Health Phys 62,.258-264 (1992).
[19]. BERAL V, FRASER P, CARPENTER L, BOOTH M, BROWN A, ROSE G.. Mortality of
employees of the Atomic Weapons Establishment, 1951-1982. Brit Med J 297:
757-770 (1988).
[20]. KENDALL GM, MUIRHEAD CR, MACGIBBON BH, O'HAGAN JA, CONQUEST AJ, GOODILL
AA, BUTLAND BK, FELL TP, JACKSON DA, WEBB, MA, HAYLOCK RG E, THOMAS JM, SILK TJ.
Mortality and occupational exposure to radiation; fist analysis of the National
Registry for Radiation Workers. BMJ 23 January 1992: 220-25 (1992).
[21]. DARBY SC, DOLL R. Radiation and exposure rate. Nature 344:824 (1990).
[22]. WALDREN C, CORREL L, SOGNIER A, PUCK TT. Measurement of low leveis of
x-ray mutagenesis in relation to human disease. Proc Natl Acad Sci USA 83:
4839-44 (1986)
[93]. GROSOVSKY AJ, LITTLE JB. Evidence for linear response for the induction of
mutations in human cells by x-ray exposures below 10 rads. Proc Nati Acad Sei
USA 82: 2092-2095 (1986)
[24]. LITTLE JB. Low-Dose Radiation Effects: Interactions and Synergism, Health
Phys 59: 49-56 (1990)
[25]. LI CY, YANDELL DW, LITTLE JB. Evidence for coincident mutations in human
lymphoblast clones selected for functional loss of a thymidine kinase gene.
Molecular Carcinogenesis, in print, 1992.
[26]. CARNES BA, FRITZ TE. Responses of the beagle to protracted irradiation.
Rad Res. 128:125-32 (1991)
[27]. PETKAU A. Radiation carcinogenesis from a membrane perspective. Acta
Physiol Scand, Suppl. 492:81-90 (1990)
[28]. GARDNER MJ, SNEE MP, HALL AJ, POWELL C.A, DOWNES S, TERRELL JD. Results of
case-control study of leukaemia and lymphoma among young people near Sellafield
nuclear plant in West Cumbria. Br Med J 300: 423-434 (1990)
[29]. KHADIM MA, MACDONALD DA, GOODEAD AD DT, LORIMORE SA, MARSDEN SJ and WRIGHT
EG. Transmission of chromosomal instability after plutonium irradiation. Nature
356: 738-740 (1992) and EVANS HJ. Alpha-particle after effects. Nature 355.
674-675 (1992).
[30]. LITTLE JB, GORGOJO L, VETROVS H. Delayed appearance of lethal and specific
gene mutations in irradiated mammalian cells. Int J Radiation Oncology Biol Phys
19: 1425-29 (1990).
[31]. CHANG WP, LITTLE JB. Persistently elevated frequency of spontatieous
mutations in progeny of CHO clones surviving X-irradiation: "Association
with delayed reproductive death phenotvpe". Mutation Research 270: 191-199
(1992).
[32]. VICKER MG, BULTMANN H, GLADE U, HÄFKER T. lonizing radiation at low doses
induces inflammatory reactions in human blood. Rad Res 128: 251-7 (1991).