|
Items in cart [1] |
Questions? Call 888-624-8373 |
| Copyright © 2009. National Academy of Sciences. All rights reserved. Terms of Use and Privacy Statement |
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 303
7
Transuranic Elements
INTRODUCTION
Transuranic elements are members of the actinide series beyond
uranium, beginning with neptunium (atomic number 93~. The last in
the series is element 103 (lawrencium). All are artificially produced
in nuclear reactors, accelerators, or explosions of nuclear weapons,
and ad have several isotopes that emit alpha rays. The energies
of the alpha particles emitted from the transuranic elements range
from about 5 to well over 8 MeV, with the higher-energy alphas com-
ing largely from the isotopes with the shortest half-lives. Berkelium
(element 97), einsteinium (element 99), fermium, mendelevium, no-
belium, and lawrencium are produced in such small amounts, mostly
for research purposes; and most of the isotopes produced have such
short half-lives, a few seconds or minutes, that they are an unlikely
health concern. Californium (element 98), a useful neutron radiation
source, is available in slightly larger amounts. Neptunium, pluto-
nium, americium, and curium, elements 93 to 96, respectively, are
the most abundant and the most extensively used of these man-made
actinide series elements. All are produced in nuclear reactors and,
because of the alpha-ern~tting isotopes with very long half-lives, for
example, 2.1 x 106 yr for 237Np, 24,400 yr for 239 Pu, 458 yr for
24iAm, and 17.6 yr for 245Cm, comprise a major radioactive waste
disposal concern. Table 7-1 lists the principal transuranic elements
which constitute potential health hazards.
303
OCR for page 304
304 HEALTH RISKS OF RADON AND OTHER ALPHA-EMITTERS
TABLE 7-1 Transuranium Nuclides of
Potential Biological Significance
Mean or Energy
Element Isotope Half-Life (yr) (MeV)
Neptunium 237Np 2.1 X 106 4.7
Plutonium 23spu 86 5.6
239pu 24,400 5.2
24opu 6,580 5.3
24~pu 13 5.1
242pU 379,000 5~0
Americium 241 Am 458 5.6
243Am 7.370 5~4
Curium 242cm 0.45 6.0
244cm 17.6 5.9
Californium 252cf 2.7 6. 2a
Einsteinium 252Es 0.5 6.7
UIncludes energy of fission fragments, neutrons, and gamma
rays.
Plutonium-239 is a constituent of nuclear weapons and, since 5
metric tons were dispersed into the atmosphere and the environment
by the nuclear weapons tests of the l950s and 1960s, trace amounts
can be found almost everywhere. Since relatively large amounts are
present in nuclear power reactors, the potential release of plutonium
in a reactor accident is a concern, although none was released in the
Three Mile Island accident and only small amounts were released at
ChernobyI.42 Plutonium-238, with a half-life of 86.4 yr, is 280 times
more radioactive per unit mass than 239Pu. Because of this high
specific activity, it is used as a heat source to power thermoelectric
devices used in cardiac pacemakers and space vehicles. The use of
very small amounts of 238 Pu in pacemakers has not caused concern,
but the potential for the reentry and destruction of space vehicles,
dispersing kilogram quantities of 238 Pu into the environment, is a
concern. Americium-241 is a contaminant of plutonium in nuclear
weapons and thus has been distributed throughout the environment.
Very small amounts have become a dependable source of ionizing
radiation required in battery-powered smoke detectors. This use has
not caused public health concern. Relatively large quantities of 237Np
are produced in fission reactors and, with plutonium, americium, and
curium isotopes, must be dealt with as a contaminant in cooling wa-
ter and as a long-lived component of nuclear reactor waste. For all
OCR for page 305
TRANSURANIC ELEMENTS
305
of the transuranic elements, occupational exposures pose a greater
potential for causing detectable health effects than environmental ex-
posures, but there is a greater potential that much larger populations
will be exposed to it by environmental exposures, but only to trace
amounts. Occupational exposures at nuclear materials production
facilities have resulted from inhalation of airborne transuranic ele-
ments accidentally released from containment equipment and from
entry through wounds occurring in the hands of persons handling
these materials in glove boxes.
The biological effects of plutonium and other alpha-emitting
transuranic elements, unlike gamma-emitting radionuclides, are pri-
marily dependent upon their entering the body and being deposited
in radiosensitive tissues. Further, the presence of transuranic ele-
ments in the environment does not necessarily infer their deposition
in human tissues. In the following discussion it will be seen that
transuranic elements are not readily absorbed from the gastrointesti-
nal tract and are even less readily absorbed through intact skin. If
environmental conditions lead to transuranic elements becoming air-
borne, there is a chance of their being inhaled and deposited in the
respiratory tract. Deposition in the respiratory tract represents the
highest probability for eventual health effects.
This chapter describes the disposition of transuranic elements
that enter the body and the biological effects that may result; it
also discusses methods for estimating risks and suggests estimates of
risk derived from other sources that might be applied to transuranic
elements in human beings. Nearly all of the information on health
effects has come from laboratory experiments since there are few hu-
man data. However, there are human data to supplement extensive
animal data on the distribution of transuranic elements in the tissues
of the body. For example, since the beginning of the Manhattan
Project in 1943, from 5,000 to 10,000 persons have been employed in
positions in the United States involving risk of plutonium exposure.
Follow-up of the distribution of plutonium in tissues has been ac-
complished by obtaining tissue samples at autopsy, or infrequently,
from surgical specimens from persons who received such exposure.
By 1986, the U.S. Iransuranium Registry has collected data from
about 200 autopsies on exposed workers whose tissues showed in-
creased concentrations of plutonium. Elevated concentrations of
plutonium in tissues of the general population are attributed to fall-
out from atmospheric nuclear weapons testing during the period from
1945 to 1963.~9
OCR for page 306
306 HEALTH RISKS OF RADON AND OTHER ALPHA-EMITTERS
Since this is not intended as an exhaustive or definitive review ot
the subject, no attempt is made to ensure comprehensive or specific
documentation of the information presented. It is intended, however,
that all information can be traced to its source through the literature
cited, especially the several reviews and symposia publications from
which much of the information was obtained.34364~7680~7~426
Although the committee intended to use only information published
in the open literature, reference is made to several recent highly
relevant laboratory annual reports.
ROUTES OF INTAKE AND DEPOSITION IN THE BODY
PERCUTANEOUS
Because of the relatively short range of alpha particles in tis-
sues, the radiation-sensitive cells of the basal layer of the skin are
not irradiated unless the alpha-emitting radionucTides penetrate the
stratum corneum or horny layer of the epidermis. The unbroken
skin has been shown to be an effective barrier to the penetration of
transuranic elements. This has been observed in skin contamination
incidents in nuclear industries and in animal experiments.4' 46 47 8~ 82
Insoluble forms such as oxides are easily removed from intact
skin by washing. Soluble transuranic compounds, such as nitrates, ci-
trates, chlorides, and complexes with organic solvents, have a greater
potential for absorption, even though it is very small. The most
common human skin exposures involve plutonium nitrate in nitric
acid solutions, plutonium tributy~phosphate in carbon tetrachIoride,
and plutonium in hydrochloric acid. These can be effectively washed
from skin with chelating compounds or detergents.
If plutonium, americium, neptunium, or einsteinium is deposited
on human or animal skin in a wide range of nitric acid concentrations
(0.1 to 10 N) for 1 h, about 5 x 10-4 is absorbed. Americium nitrate
in tributy~phosphate exhibits nearly a factor of 10 greater percu-
taneous absorption, about 3 x 10-3.46 While data for transuranic
oxides are sparse, one can predict that percutaneous absorption
through intact skin would be less than 1 x 10-5 during the first
hour after deposition. An approximate Unfold increase in absorp-
tion was seen when soluble transuranic compounds remained on the
skin for 3-5 days; over this prolonged time, there was evidence that
higher concentrations of nitric-acid-enhanced percutaneous absorp-
tion. Under the most extreme conditions, such as plutonium nitrate
OCR for page 307
TR-ANSURANIC ELEMENTS
307
in 10 N nitric acid for 4 days or americium nitrate in 8 N nitric acid
for 3 days, maximum percutaneous absorption was only 2%.
Damage to skin by trauma, wounds, acid, or thermal burns
facilitates a more rapid transfer of soluble transuranic compounds
into the subcutaneous tissue and blood. Insoluble particles and
metal slivers deposited below the level of the epidermis are slowly
cleared to regional lymph nodes.
There are three mechanisms for transport of transuranic ele-
ments from skin: (1) transfer into the subcutaneous microcirculation
and then into the blood and lymphatic systems, (2) transfer onto the
skin surface with sudoriferous and sebaceous secretions, and (3) loss
from body with desquamation of skin. Autoradiographic studies of
cutaneously deposited plutonium, americium, and neptunium show
a decreasing concentration with increasing depth in skin, but with
focal concentrations of activity in the upper epidermis, hair follicles,
sebaceous glands, and microvasculature.47
In the event of accidental occupational exposure to plutonium
through skin wounds in the hands and other sites of the body, plu-
tonium may be retained at the wound site and removed by surgery,
sloughed from the surface, solubilized and transiocated to internal
organs, or transported to regional lymph nodes. Wounds penetrating
the horny layer of skin lead to more rapid absorption and transio-
cation of soluble transuranic compounds to bone, liver, and other
tissues.
The most serious of these accidents, in terms of quantity of ac-
tivity deposited in the body, involved an explosion that resulted in
deposition of a total of 1-5 mCi of 24tAm on the face and, by ~nhala-
tion, in the respiratory tract. About 1 mCi remained in the body,
mostly in wounds on the face, after initial emergency decontami-
nation. Chelation therapy by intravenous injection of diethylene-
triaminepentaacetic acid (DTPA) facilitated the excretion of 900
Foci.
Investigators gave beagle dogs subcutaneous implants of 9.5 psi
239 PuO2 or 1.3 psi 239Pu(NO3~4 in their forepaws to mimic hand
wounds received by plutonium workers. At 5 and 8 yr following
exposure, the injected paws still retained 21 and 16%, respectively,
of deposited plutonium; in both cases the highest concentrations of
activity were found in regional lymph nodes, with the liver showing
the next highest concentration of plutonium. The skeleton retained
a greater amount of plutonium from the nitrate than from the diox-
ide compound.25 Similar results were observed with subcutaneously
OCR for page 308
308 HEALTH RISKS OF RADON AND OTHER ALPHA-EMITTERS
injected monomeric and polymeric 239 Pu in mice, with the monomer
behaving like plutonium nitrate and the polymer like plutonium
dioxide.44
GAsTRoINTEsTINAL TRACT
The gastrointestinal tract, provides a substantial barrier to the
uptake of transuranic elements ingested with food or water. Al-
though the fraction absorbed is usually tow, continuous ingestion of
contaminated food and water may lead to the~presence of measurable
amounts in the body. Gastrointestinal absorption is also a consider-
ation in assessing the risk from inhaled transuranic elements because
of clearance from the lungs to the gastrointestinal tract, but is small
compared to direct absorption from the lungs.
The importance of chemical form and oxidation state in the
absorption of transuranic elements has been verified in several labo-
ratories. These data have been tabulated by the International Com-
mission on Radiological Protection (ICRP).4t Relatively high con-
centrations of chlorine in some domestic water supplies could oxidize
Pu+4 to Pu+6.52 However, subsequent animal studies failed to pro-
vide convincing evidence that changes In valence state for ingested
plutonium had a significant effect on gastrointestinal absorption.
Fasting may increase the absorption of plutonium by about 1 order
of magnitude. Both calcium- and iron-deficient diets tend to en-
hance absorption of plutonium.87 ti2 The absorption of transuranic
elements incorporated in food fed to experimental animals may be
2 to 10 times greater than the absorption of chemical forms such as
citrate and nitrate.
In adult animals, <0.01~o of plutonium and most other transura-
nic elements is absorbed from the intestines.4i Overall, there is lit-
tle variation in absorption of plutonium and americium nitrates in
adult rats, guinea pigs, or dogs.~° Swine exhibit a greater plutonium
absorption than rats or dogs. The absorption of 237 Np can be in-
creased 10 to 100 times by increasing the mass ingested; however,
at occupationally and environmentally relevant levels the absorp-
tion is more like that of the other transuranic elements. With few
exceptions, the absorption of transuranic compounds from the gas-
trointestinal tract in adult experimental animals varies over 3 orders
of magnitude 10-5 to 10-2. This led the ICRP4t to adopt values
of 0.1 x 10-4 for plutonium oxides and 1 x 10-4 for plutonium ni-
trate for application to occupational exposures. These values would
OCR for page 309
TR24NSURANIC ELEMENTS
309
apply to inhaled material cleared from the respiratory tract to the
gastrointestinal tract, as well as to ingested material. For all other
plutonium compounds and compounds of all other transuranic ele-
ments, including those incorporated in food products and drinking
water, the ICRP adopted an absorption factor of 10 x 10-4 for
application to exposures of both workers and the public.*
Ingestion of alpha~emitters is not considered a radiological haz-
ard to the gastrointestinad tract since the range of alpha particles is
insufficient to penetrate the mucus and intestinal contents and reach
the crypt cells.
The neonatal rat and guinea pig absorb about 100 times more
plutonium than the adult, while newborn swine absorb 20 times more
plutonium than either the newborn rat or guinea pig. In addition to
the increased fraction absorbed, a substantial fraction of transuranic
elements given in soluble form was retained for several days within the
mucosa of the small bowel.~09 The ICRP has adopted an absorption
value of 100 x 10-4 for the first year of life in contrast to the value
of 10 x 10-4 cited above for all later years.4t
RESPIRATORY TRACT
Inhalation is probably the most common pathway by which
transuranic elements cross the barriers of the body and penetrate
into and across living cells. The aerodynamic particle size of the
aerosol, which accounts for not only the sizes of particles but also
their density and shape, determines the fractional deposition and
sites of deposition in the respiratory tract. The subsequent rates and
routes of clearance; the transIocation to, deposition in, and rate of
clearance from other tissues; and the excretion in urine and feces
of inhaled transuranic compounds depend on particle size, solubil-
ity, density, shape, and other physicoche~cal characteristics of the
aerosol. In this way the physical and radiological properties of the
transuranic compound, and the physiological characteristics of the
exposed individual determine the amount deposited and thus, the
radiation dose rates and total doses delivered to the tissues of the
*These are revised from those used to calculate annual limits on intake by
the ICRP.4i The fat values (fraction transferred to blood) used were 0.1 x 10-4
for plutonium oxides; 1 x 10-4 for nitrates and other plutonium compounds;
5 X 10-4 for all americium, curium, and californium compounds; and 100 x
10-4 for neptunium compounds.
OCR for page 310
310 HEALTH RISKS OF RADON AND OTHER ALPHA-EMITTERS
respiratory tract and other organs of the body. Aerodynamic parti-
cle diameter is a useful predictive characteristic of an aerosol for the
estimation of deposition in regions of the respiratory tract. Several
dosimetric models have been developed for describing particle depo-
sition and clearance in the human respiratory tract. These models
provide a basis for estimating deposition, distribution, and reten-
tion of inhaled radioactive aerosols, taking into account particle size
and chemical form of the aerosol. Mathematical models were devel-
oped to describe the deposition and clearance of inhaled materials
from the several compartments of the nasal passage, the trachea and
bronchial tree, the pulmonary parenchyma, and the thoracic lymph
nodes. The models, when used for radiation protection purposes,
apply to a reference man, a 70-kg male worker. Thus, they can
be expected to only approximate the deposition, distribution, and
retention of inhaled radionuclides in any given individual.
For inhalation of an aerosol with an activity median aerody-
namic diameter (AMAD) of 1 ,um, according to the ICRP mode}
(based largely on experiments with nonradioactive aerosols in human
subjects), 30% of the particles are deposited in the nasal passages,
Who in the trachea and bronchial tree, and 25% in the puirnonary
parenchyma, for a total deposition in the respiratory tract of 63~o
of the amount inhaled. The amount exhaled, not deposited, ~ 37%.
These deposition fractions will vary with particle size and the breath-
ing rate and volume.39 The ICRP has devised metabolic models to
describe the retention and transIocation of transuranic compounds
from these sites of deposition, based largely on data from experimen-
tal animals. The following summarizes this information for several
transuranic compounds.
Following deposition in the lungs, particles are quickly phagocy-
sized by alveolar macrophages. The attenuated cytoplasm of type
alveolar epithelial cells may also phagocytize particles.95 Up to loo
of particles, including transuranic oxides, deposited in the Jung may
also be taken up by tracheobronchial and bronchiolar epithelia.l3
Particles penetrating the respiratory epithelium may be phagocy-
tized in interstitial areas and, if insoluble, eventually cleared to
regional lymph nodes of the thoracic cavity. Since relatively soluble
tra~suranic compounds, such as nitrates, citrates, and the oxides of
americium and curium, are rapidly cleared into the blood, only small
fractions are cleared to lymph nodes. However, plutonium inhaled as
relatively insoluble plutonium oxide particles is very slowly cleared
into the blood. Thus, within a few hours after inhalation, about half
OCR for page 311
TRANSURANIC ELEMENTS
311
of inhaled 239 PuO2 deposited in the alveoli can be removed from ex-
perimental animal lungs by multiple lavage,7t 92 with most particles
having been phagocytized by pulmonary macrophages. This phago-
cytosis of plutonium particles may facilitate their transport from the
lungs by the mucociliary epithelium and possibly contributes to their
transport to lymphatic tissues.
The distribution of transuranic elements within the lungs is rel-
atively uniform after inhalation, more uniform for the most soluble
forms. Following the initial clearance process and especially the am
sorption of the more transportable material by the blood circulating
through the lungs, the distribution within the lungs becomes much
less uniform. The material retained in the lungs for long times is lo-
calized primarily in bronchiolar, alveolar, and lymphatic structures
of the lung parenchyma, frequently in regions of fibrosis and scar
tissue.24 This pattern appears to be consistent among all experimen-
tal animals studied, and while there are few observations, there is no
evidence to the contrary for human lungs.
Respiratory tract clearance of inhaled plutonium in human ac-
cidental exposure cases is similar to that seen in PuO2 studies
with large animals (dog,5 70 sheep,~°8 baboon,8 burro,' °8 and rhesus
monkey50) with half-times for three exponential phases of approxi-
mately 1, 30, and 300 to over 1,500 days, respectively.424 The second
phase is not always distinguishable. Early clearance of plutonium is
from the nasal passages and upper tracheobronchial regions, while
clearance with longer half-times is from the bronchiolar and alveolar
or pulmonary regions. High fecal to urine 239Pu ratios (between 50
and 500), indicative of a high insolubility, are observed in humans
for long periods following inhalation of 239PuO2. A large fraction, as
much as 50~o, of inhaled and deposited insoluble 239PuO2 and up to
about 25% of 238 PuO2 may eventually be transported to the thoracic
lymph nodes of ~ogs.83
In contrast to 239Pu02, inhaled 24iAm and 244 Cm dioxides, as
well as plutonium nitrates, in humans and animals are relatively
soluble, with about half of the amounts deposited in the bronchiolar
ant] alveolar regions cleared with a half-life of 10 to 40 days and the
remainder cleared with half-lives generally ranging from 200 to 500
days. Less than l$to of these relatively soluble transuranic compounds
deposit in thoracic lymph nodes.6
While plutonium oxide particles are generally quite insoluble
in the respiratory tract, there are some exceptions. For example,
OCR for page 312
312 HEALTH RISKS OF RADON AND OTHER ALPHA-EMITTERS
it has been demonstrated in several animal species that the con-
ditions under which plutonium is oxidized may affect the fate of
particles deposited in the respiratory tract.5 Electron micrographs
suggest that plutonium particles oxidized at high temperatures have
less surface area than those oxidized at much lower temperatures
and, thus, could have lower dissolution rates in body fluids. This
was verified in studies in which plutonium oxide particles formed at
high temperatures (over 1,000°C) tended to have lower transIoca-
tion rates from the lungs than plutonium oxidized in air at ambient
temperatures or calcined at relatively low temperatures. Also, alve-
olar clearance and transiocation of 233PuO2 to other tissues such as
liver and bone are nearly always more rapid than those for com-
parably prepared 239PuO2.56 83 Plutonium-238 is 280 times more
radioactive than an equal mass of 239Pu. Radiolysis may cause these
high-specific-activity 238PuO2 particles to fragment within the lungs,
greatly increasing the surface area of the 238Pu particles, and thus
their dissolution rate.
Nanometer-diameter plutonium oxide particles have been found
to be cleared from the respiratory tract very rapidly and appear to
be excreted in the urine as particles.~05 If transuranic elements are
inhaled simultaneously with other materials, their disposition may
depend on how the transuranic element is combiner} with the other
material in the aerosol. For example, calcining 239PuO2 with a rel-
atively large amount of sodium, potassium, calcium, aluminum, or
uranium increases the solubility of 239Pu in the lung.3 306 Increas-
ing the ratio of plutonium to sodium in laser-vaporized aerosols of
PuO2-UO2 and sodium from 0 to 1:1 and to > 1:10 increased the
rate of clearance from the lungs and transIocation to extrapulmonary
tissues from O.S to 5.0 and 24~o, respectively.53 After inhalation of an
aerosol of 239 PuO2 and 244CmO2 calcined as a mixture, both pluto-
nium and curium remained in the lung somewhat longer than when
caTcined and inhaled separately.~°i The transiocation of curium to
extrapulmonary tissues was largely prevented by incorporation into
the much greater mass of the PuO2 matrix. However, in rats the rate
of alveolar clearance and transIocation of i69Yb and 239 Pu inhaled
as an oxide, prepared by calcining i69Yb mixed with 239 Pu, were not
significantly different from the rates of clearance and transIocation
of i69Yb2O3 or 239 PuO2 inhaled separately.~02
The high rate of accumulation of inhaled insoluble plutonium
in lymph nodes has stimulated considerable interest. Lymph nodes
draining the lungs attain concentrations of inhaled plutonium many
OCR for page 313
OCR for page 356
OCR for page 357
OCR for page 359
OCR for page 360
OCR for page 361
OCR for page 362
OCR for page 363
OCR for page 364
OCR for page 365
OCR for page 366
Representative terms from entire chapter:
lymph nodes
TRANSURANIC ELEMENTS
313
times higher than those of any other tissue. The particles are pref-
erentially localized along sinusoids in the paracortical area and in
medulIary cords and less so in the lymphoid germinal centers.40
More than 10% of 239PuO2 deposited in the alveoli was taken up
by thoracic lymph nodes of dogs by 1 yr postexposure, increasing
to 15~o by 2 yr and 30 to So-so by 5 to 15 yr. Accumulation of
inhaled 238PuO2 in thoracic lymph nodes was less than 239PuO2; it
reached a maximum of 20 to 24~o and gradually declined to
356 HEALTH RISKS OF RADON AND OTHER ALPNA-EMITTERS
TABLE 7A-3 Summaries of the Individual Studies
Isotopea
Biological System 226Ra 228Ra 23spu 239pu
Human —3.30 (0.32) - 2.43 (0.72) — —
Beagle dog (injection) - 0.87 (0.21) - 0.32 (0.16) — 1.69 (0.17)
Beagle dog (inhalation) — — 1.55 (0.15)
Rat — - — 0.64 (0.34) 0.29 (0.24)
NOTE: Dashes in this table correspond to combinations of isotope and biological system with no
available study. In case data from multiple studies were available for a cell, estimates were aver-
aged, as if they were unbiased estimates of the same quantity, to produce one value of y and c in
each such cell. The weight given to each y was inversely proportional to the square of its value of
c, and values of c-2 were summed to obtain the value of c-2 for the average.
aData are arranged as follows: Yij, in units of log tumors/1,000 reds (cij, as estimated standard
error of Yij)-
system-isotope combination. Conditional on bit, Is is assumed to be
normally distributed with standard deviation cij.Formally:
Yij~6ij ~ N(8ij~cij2~.
(7A-2)
The values of ~ have normal prior distributions that are independent,
conditional on the values of further parameters:
off,. . .a4,7l,. · .14, Ed a:
ij~, 1, a) ~ N(ai + A, ~2~.
(7A-3)
Finally, the parameters se, A, and ~ all have prior distributions
as discussed below. The specification of these prior distributions,
together with the values of y and c given in Table 7A-3, completes
the specification of the formal Bayesian model. Equations given by
DuMouche! and Harrist then provide final estimates and standard
deviations for each of the bid values, including those for which no
corresponding Yij value is available.
The crux of the Bayesian analysis is the specification of the prior
distributions for a!, I, and a. In order to do this, it is necessary
to have a good understanding of the meanings of these parameters
and their conceptual role in the analysis. These parameters, which
are only used to specify the prior distribution of the bij term, the
parameters of direct interest, are called hyperparameters.
A BAYESIAN METHODOLOGY FOR COMBINING RADIATION STUDIES 357
INTERPRETATION OF THE HYPERPARAMETERS
The prior mean of bit, the log of the slope of bone-cancer risk
versus dose, is, by Equation 7A-3, the sum of an average for the ith
biological system, se, and an effect due to the jth isotope, Hi. This
additive model translates to a multiplicative model on the original
scale. This means that the prior expectation is that the ratio of
the carcinogenic potency of any two isotopes is preserved across
species. The hyperparameter ~ measures how well the actual his
values conform to this prior expectation. A belief that a is very near
zero implies a belief that the relative potency of isotopes is almost
exactly the same for every species. Larger values of or imply more
probability that some of the species systems have isotope-specific
reactions to radiation.
The fact that the mean value of each aft has an additive rep-
resentation implies that we cannot identify a priori which biological
system is most likely to exhibit a distinctive reaction to any par-
ticular isotope. The specific values of pi measure the sensitivity of
the ith biological system to the average isotope, while the specific
values of the hi measure the average potency of the jth isotope across
biological systems.
PRIOR DISTRIBUTIONS USED IN THE ANALYSES
DIFFERENCES BETWEEN BIOLOGICAL SYSTEMS
The values of Cal, 0~2, are, old, respectively, represent the average
log potency of the isotopes being considered, when dose is measured
as reds to the skeleton, in the four biological systems. Except for
the data now being analyzed, there is very little knowledge about
these quantities. Therefore, it seems appropriate to choose prior
distributions for the a, term that are very broad. These parameters
are all assumed to have identical normal distributions, with a mean
of 0 and a standard deviation of 10. (Note that all of the values
of y are between -4 and +2. This shows that the data are much
more precise than the assumed marginal prior distributions of pi.)
Although there is little prior information about the individual
358 HEALTH RISKS OF RADON AND OTHER ALPHA-EMITTERS
pathway to the skeleton than ~ either ingestion or injection. The
ingestion and injection experiments are designed to be as comparable
as possible, and theoretically, their bone-cancer effects should be the
same if the dose to the skeleton is the same. However, considering the
difficulty in determining dose to the skeleton, metabolic differences,
and other differences between the two groups of beagles, one cannot
rule out the possibility that all of these differences together result
in a systematic difference in the estimated potency of all isotopes.
This possibility will be described by following the prior probability
statement:
Pow - aid > 0.2) = 0.05.
(7A-4)
This states that there is only a 5%o probability that the ratio
of average potencies (tumors per red to the skeleton) from the two
modes of administration is greater than e0 2 in favor of either mode.
In terms of the assumption of normal distributions, this translates to
an assumption that the standard deviation of a2 - ore is 0.1. Since
the variances of each c'' have been assumed to equal 100, this implies
that the covariance of °~2 and of is 99.995. To summarize, the prior
distributions of the pi are assumed to be normal with means of O and
· —
covar~ance matrix:
100 0 0 0
0 100 99.995 0
0 99.995 100 0
0 0 0 100
The fact that the variances and covariances are chosen to be
exactly 100 and 99.995 is not crucial here. The only important
feature is that the standard deviation of a2 - Ct3 is 0.1, and the
standard deviations of all other linear combinations of ai are assumed
to be very large. Any other prior distributions of pi that have these
features lead to almost exactly the same results.
DIFFERENCES BETWEEN ISOTOPES
Next, the prior distributions of hi are considered. These distri-
butions represent the average differences (across species), on a log
scale, of the potencies of the four isotopes under consideration. Here
there is some scientific knowledge. If one really believed ha red is a
red is a red," then one would assume that every as = 0. However,
the possibility that the different isotopes have different potencies per
A BAYESIAN METHODOLOGY FOR COMBINING RADIATION STUDIES 359
red to the skeleton wiD be assumed here. The hyperparameters 7,
and 72 correspond to the isotopes 226 Ra and 228Ra, respectively. The
latter isotope has a much shorter half-life, a different decay chain
with daughters that emit different radiations of different energies,
and this may interact with the phenomenon of carcinogenesis in un-
predictable ways. It is barely possible that either of the two radium
isotopes is as much as twice as potent as the other. This will be
stated probabilistically as:
P(~7~ - 72~ > logic = o.o5.
(7A-5)
Similarly, 73 and 74 correspond to the isotopes 238 Pu and 239Pu,
respectively. These two isotopes each have very long half-~;ves, and
it is harder to find a rationale for the possibility of a consistent dif-
ference between these two isotopes. Accordingly, a ratio of potencies
of 1.5 is barely possible here. Probabilistically,
P(,~73 - 74~ > log 1.5) = 0.05.
(7A-6)
Finally, compare the potencies of radium and plutonium. Here
there is also scientific knowledge. Because it is known that plutonium
concentrates more in the outer layers of bone ceils than does radium,
and because osteosarcomas also tend to originate in these layers of
cells, the same dose to the skeleton of plutonium will tend to produce
more tumors in all species than wiD radium. The relative potency of
either isotope of plutonium to either isotope of radium is judged to
be almost surely greater than 1 but less than 10. Probabilistically,
P(O < 7j—Ok < log 10) = 0.95; j = 3, 4 and k = 1, 2. (7A-7)
If one uses the assumed normality of the prior distributions of
A, the above probabilities can be used to derive the means and
covariance matrix of A. The means are (-0.25 log 10, -0.25 log 10,
0.25 log 10, 0.25 log 10~. The prior covariance matrix for the By is:
0.166
0.106
0.0
0.0
0.106
0.166
0.0
0.0
0.0
0.0
0.166
0.145
0.0
0.0
0.145
0.166
Using the terminology of DuMouche! and Harris, the values of
Y. C, X, i, and V are now specified for the Bayesian analysis. The
values of Y. C, and X are given in Table 7A-4.
360 HEALTH RISKS OF RADON AND OTHER ALPHA-13MITTERS
TABLE 7A-4 Values of Y. C, and X
Biological System Isotope Y C X
Human 226Ra —3.30 0.32 1 0 0 0 1 0 0 0
Human 228Ra —2.43 0.72 1 0 0 0 0 1 0 0
Beagle dog (injection) 226Ra —0.87 0.21 0 1 0 0 1 0 0 0
Beagle dog (injection) 228Ra —0.32 0.16 0 1 0 0 0 1 0 0
Beagle dog (injection) 239pu 1.69 0.17 0 1 0 0 0 0 0 1
Beagle dog (inhalation) 23epu 1.55 0.15 0 0 1 0 0 0 1 0
Rat 238pu 0.64 0.34 0 0 0 1 0 0 1 0
Rat 239pu 0.29 0.24 0 0 0 1 0 0 0 1
TABLE 7A-S Values of b and V
Prior
Mean
Hyperparameter b
Human
Beagle dog (injection)
Beagle dog
(inhalation)
Rat
226Ra
228Ra
238pu
239pu
Prior Covariance Matrix V
o.o 100.0 o.o o.O 0.0 0.0 0.0 0.0 0.0
o.o o.o 100.0 99.99s o.o o.o o.o o.o o.O
o.o o.o 99.995 100.0 o.o o.o o.O 0.0 0.0
o.o o.o
—o.s8 0.0
-0.s8 o.o
o.s8 o.o
o.s8 0.0
o.o
o.o
o.o
o.o
o.o
0.0 100.0 0.0 0.0 0.0 0.0
0.0 0.0 0.166 0.106 0.0 0.0
0.0 0.0 0.106 0.166 0.0 0.0
0.0 0.0 0.0 0.0 0.166 0.14S
0.0 0.0 0.0 0.0 0.14S 0.166
The values of C in Table 7A-4 are estimated standard errors.
They would be squared and then represented as a diagonal matrix
to conform to the notation of DuMouche] and Harris.1 The first four
columns of X identify the four biological systems, while the last four
columns of X identify the four isotopes. The corresponding values of
b and V are given in Table 7A-5.
~ ,
PRIOR DISTRIBUTION FOR Cr
r~' ~ ~ ~
v -
~ he value of ~ determines how reliable the interspecies extrapo-
lation is expected to be. From Equation 7A-3 each log potency' bij'
has prior mean ai + 7j and prior standard deviation ~. For any two
biological systems, i and il. and any two isotopes' j and ji, the linear
combination ~ = ~ij - ~ilj - ~ijl + biljl = log (Aij/Ailj)/(Aijl/Ai~jl),
is assumed to be normally distributed with mean of O and a standard
deviation of 2a, conditional on a. The interpretation of 2\ is that
e'` is the ratio by which the extrapolation of potency fails when the
A BAYESIAN METHODOLOGY FOR COMBINING RADIATION STUDIES 361
isotopes j and j' are compared for the pair of biological systems i and
i', if individual potencies were perfectly measured. We judge that
this extrapolation is highly unlikely to fail by more than a factor of
10. Probabil~stically,
PA > log 10) ~ 0.05.
Now, conditional on a:
(7A-8)
PA > log 10a) = 2~1—Clog 10/2~], (7A-9)
where ~ is the standard normal distribution function. Therefore,
PA > log 10) = E{2~1 - Clog 10/2~}, (7A-10)
where E{) refers to the expectation with respect to the prior distri-
bution of a. If we assume that, a priori, ~ takes each of the 10 values
0.05, 0.15, . . ., 0.95 with a probability of one-tenth, then expectation
in Equation 7A-10 is, in fact, about 0.06, which is in agreement with
the subjective assessment of Pit ~ ~ ~ log 10~. Therefore, this prior
distribution for a is used in the Bayesian analysis.
RESULTS OF THE BAYESIAN ANALYSIS
Having defined the quantities Y. C, X, b, V, and the prior distri-
bution of a, it is now straightforward to use the procedures given by
DuMouche} and Harris t to compute the posterior distributions of or
and A. When this is done, the posterior distribution of ~ is iota LYE,
given by:
~ = 0.05 0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95
x(~Y) = 0.241 0.212 0.166 0.122 0.087 0.062 0.043 0.030 0.021 0.015
Thus, although the prior distribution of ~ was approximately
uniform over the interval (0,1), the posterior mean of ~ is 0.25 and
Pit a < 0.5 ~ Y) = 0.83. Roughly, this analysis suggests that ~ is
about half as large as was supposed a priori. Extrapolation on the
basis of the comparison of two isotopes on each of two species is likely
to be off by a factor of 3 to 5 rather than by a factor of 10. The
posterior probability that a new extrapolation will be off by a factor
of 10 or more is Pt ~ ~ ~ > log 10 ~ Y) = 0.18, down from the value
of 0.06 computed from the prior distribution.
362 HEALTH RISKS OF RADON AND OTHER ALPHA-I~MITTERS
TABLE 7A-6 Summary of Posterior Distributions after
Combining Studies
Parameter of Isotope
Posterior Distribution 226Ra 228Ra 238Pu 239pu
E{~} —3.22 - 2.81 - 1.11 - 1.12
Standard deviation {~} 0.30 0.46 0.65 0.65
Ohs 0.04 0.06 0.33 0.33
\0.02s 0.02 0.03 0.09 0.09
~o.s7s 0.07 0.15 1.12 1.12
Table 7A-6 displays information on the posterior distributions of
the potencies of each of the four isotopes in man. The last three lines
of Table 7A-6 show the medians and 95~o confidence limits of the
potency, in bone cancers per 1,000 red to the skeleton, of each of the
isotopes. These limits were computed by assuming that the posterior
distributions of ~ ~ = log A) are Gaussian, as is approximately true.
Note that the uncertainty ratios >0.975/~0.025 for the potencies of
plutonium are greater than those for radium, since no direct data on
the ejects of plutonium in man have yet been incorporated into the
analysis.
USING THE DATA ON HUMAN EXPOSURE TO PLUTONIUM
The posterior distribution for the effects of plutonium on man
resulting from the Bayesian analysis described above, in which the
data from Rowland and Durbin4 5 were not used, is now used as the
prior distribution for the analysm of those data. This second anal-
ysis proceeds as a Bayesian update by using the Poisson likelihood
function of the data, namely
L(~) = Ptn = 0~A,N= 1,D,T) = expi—ALiDi(1—r/Ti)], (7A-ll)
where D (`D,, D2, . . .) and T = (T1, T2, . . .) are, respectively,
the doses (in thousands of reds) and observation times (in years) for
the individuals included in the studies. Each individual is considered
as a group of size N = 1, and since no bone cancers were observed in
these individuals, every n = 0. The value r = 5 yr was used as the
latency parameter for this analysts. When the values of D, T. and
are substituted into the above formula for ]; Ail, it becomes:
A BAYESIAN METHODOLOGY FOR COMBINING RADIATION STUDIES 363
TABLE 7A-7 Converting from Log Normal to Gamma Distributions
Parameter of Isotope
Posterior Distribution 226Ra 228Ra 23epu 239pu
E{log X} - 3.22 - 2.81 - 1.11
Standard deviation {log \} 0.30 0.46 0.65
E{ ~ } 0.042 0.067 0.406
Standard deviation Id} 0.013 0.032 0.294
Gamma c 10.3 4.3 1.9
Gamma d 246.1 64.1 4.7
L(~) = exp( - 0~449~) (238 Pu), and
= exp( - 0~324~) (2~9pu) (7A-12)
Under our model, the expected number of bone-cancer deaths
among the individuals in the studies by Rowland and Durbin4 5 would
be 0.449> and 0.324) for those exposed to 238 Pu and 239Pu, respec-
tively. Since the previous analysis concluded that ~ is probably less
than 1, human bone-cancer death/thousand person-red of plutonium
exposure, there cannot be much further information in these data.
The previous analysis approximated the distribution of ~ by a
log-normal distribution. However, in order to combine this distri-
bution with the exponential likelihood function given above, it is
convenient to use a ga~nma-distribution approximation. For a given
mean and variance of A, the gamma distribution with the same first
two moments as the log-normal distribution will be considered equiv-
alent to it. The gamma density is:
Gyp; c, d) = tic, d)AC~texpt—Ad), (7A-13)
where c > 0 and ~ > 0 are parameters determining the mean and
variance of A, and chic,`) is a normalizing constant ensuring that the
density integrates to unity over the range ~ > 0. The mean of ~ is
c/d, while the variance of ~ is c/12. When the distributions from
Table 7A-6 are converted from log-normal to gamma representations,
the resulting values of c and ~ are shown in Table 7A-7.
If ~ has the Gig; c,`) density, the values of c and ~ have
a simple interpretation. Bayesian probability intervals for ~ then
coincide numerically with the frequentist confidence intervals which
would result if c cancers were observed in a population exposed to
a total of ct (times 1,000) red (outside the latent period). Thus, the
364 HEALTH RISKS OF RADON AND OTHER ALPHA-EMITTERS
TABLE 7A-8 Percentiles of ~
Isotope 238pu 239pu
cots 0.30 0.31
x0 025 0.08 0.09
\0.975 1.07 1.09
conclusions reached by this Bayesian analysm of the data in Table
7A-3, excluding the data on plutonium in man, are numerically very
sitar to those that might be reached by a non-Bayesian statistician
who had observed 1.9 (i.e., about 2) bone-cancer deaths in a human
population exposed to a total of 4,700 person-red from plutonium
exposure.
The Go; c,d) prior distribution is mathematically convenient
because it is easily updated: On observation of n cancers in a pop-
ulation exposed to a total of d' cumulative (latency-adjusted) thou-
sands of red, the posterior density of ~ Is Go; c + n, ~ +d'). In
incorporating the data from Rowland and Durbin,4 5 the values of n
are 0 for both isotopes, while d' = 0.449 for 238 Pu, and d' = 0.324 for
239Pu. The posterior gamma densities are therefore Gas; 1.9, 5.1)
for 238Pu and Gig; 1.9, 5.0) for 239 Pu. The mean values of ~ then
become 0.370 and 0.379, respectively. If the gamma distributions are
converted back to the log-normal distributions with the same mean
and variance, the percentiles of ~ are as given in Table 7A-8. As
can be seen by comparing the results given in Table 7A-8 with those
in Table 7A-6, the use of the human plutonium data has very little
effect on the distribution of A.
CONCLUSION AND SUMMARY
Data from several studies on the effect of internal deposition
of two isotopes of radium and two isotopes of plutonium on bone-
cancer death rates have been collected and summarized in an easily
compared form. The 15 different data tables of the quantities n,
the number of bone-cancer deaths; N. the number of individuals;
D, the total cumulative dose to the skeleton received by these indi-
viduals; and T. the total animal or person-years of observation of
the individuals, by dose group, within each study were described.
These summary statistics are often available in the published papers
that describe each study, and they are the bare minimum needed to
make any cros~study meta-analysis possible. The values n, N. D,
and T are needed for each of 5 to 10 well-spaced dose-rate groups
A BAYESIAN METHODOLOGY FOR COMBINING RADIATION STUDIES 365
within each study. It is also necessary to assume that the dose rate
is roughly constant over time and over animals within each dose-
rate group. The summary of the radium-dial painter data that were
available within the time frame of this analysis contained only three
broad dose-rate groups. With more detailed data, the analysis could
-
~e 1mprovec .
Using these summary data tables, a linear-effects dose-response
mode! was fitted to the data from each study. This produced an
estimate of the bone cancers per red observed in each study, with
an estimate of the w~thin-study sampling variation attached to each
slope. This allowed the use of a Bayesian components of variance
mode! to estimate by how much ratios of slopes from different studies
differ more than could be explained by the within-study sampling
variation. The posterior distribution of the parameter ~ showed that,
although the variation in the ratios of estimated slopes for different
isotopes cleposited in the same species could possibly be explained
purely by within-study sampling errors, making extrapolation across
species and isotopes potentially accurate, the fact that the value is
log 2 also had nonnegligible probability means that there may
be no hope of extrapolating dose-response slopes more accurately
than by a factor of 2 or 4, even if very good data on the effects
of other isotopes on human bone-cancer rates and of plutonium on
several animal systems are available. The question cannot be settled
without gathering more data from other combinations of isotopes
that act on biological systems.
In this regard, it would be greatly advisable for researchers to
consider how their proposed studies fit into the matrix of other stud-
ies already performed, so that meta-analyses of all the studies can be
most informative. For example, one crucial hole in the array of stud-
ies available was that there were no measures of the effect of radium
on bone cancer in rats. This prevented the analysis from making
effective use of the several plutonium studies on rats. Similarly, the
fact that all the radium studies on beagles used the injection mode
of dose administration, while most of the plutonium studies on bea-
gles used the inhalation mode of administration, introduced a prior
uncertainty, which lessened the accuracy of the Bayesian analysis.
The Bayesian methodology illustrated here allows a quantifi-
cation and adjustment for prior uncertainty which is impossible to
achieve by using the frequentist approach to statistical inference. The
particular prior distributions of the hyperparameters {~ ), (hi), and ~
that were used in the analyses were intended to make use of as much
366 HEALTH RISKS OF RADON AND OTHER ALPHA-EMITTERS
scientific information as possible, in addition to the information con-
ta~ned In the data under analysis.
In summary, the Bayesian analysis presented here gives all esti-
mate of the risk of bone cancer due to internally deposited plutonium
of about 300 cancer deaths per million person red for dose received
beyond the latency period of very small increased risk. The 95~o
confidence interval includes the range from about 80 to 1100 bone
cancer deaths per million person red. These risks are 5 to 10 times
larger than the estimated risks for 2269228 Ra ~ humus, but the
interval of uncertainty determined here is considered to be realistic.
Finally, the published data on a few humans injected with plutm
nium were reanalyzed and integrated into the larger analysis. It was
determined that these data are too meager to provide any important
information on the bone-cancer effects of plutonium deposition.
REFERENCES
1. DuMouchel, W. H., and J. E. Harris. 1983. B ayes methods for combining
results of cancer studies in humans and other species, with discussion. J.
Am. Stat. Assoc. 78:293-315.
2. Inhalation Toxicology Research Institute (ITRI). 1985. Annual Report
LMF-114. Albuquerque, N.M.: Lovelace Biomedical and Environmental
Research Institute.
3. International Commission on Radiological Protection (ICRP). 1976. Bi-
ological Effects of Inhaled Radionuclides. ICRP Publication 31. Oxford:
Pergamon Press.
4. Rowland, R. E., and P. W. Durbin. 1976. Survival, causes of death, and
estimated tissue doses in a group of human beings injected with plutonium
and radium. Salt Lake City: University of Utah Medical Center: The J.W.
Press.
5. Rowland, R. E., and P. W. Durbin. 1978. The plutonium cases: An
update to 1977. Oral presentation at the Scientific Group Meeting on
Long-term Effects of Radium and Thorium in Man, Geneva: World Health
Organization. (A summary can be found in pp. 138-141 in Radiological
and Environmental Research Division Annual Report, ANL-78-65, Part II,
Argonne, Ill., Argonne National Laboratory, 1978.)
6. Rowland, R. E., A. F. Stekney, and H. F. Lucas, Jr. 1987. Dose-response
relationships for female radium dial workers. Radiat. Res. 76:368-383.
7. University of Utah, Radiobiology Division. 1983. Research in Radiobiol-
ogy. Radiobiology Division Annual Report C00-119-258. Salt Lake City:
University of Utah School of Medicine.
