Introduction

Four isotopes of radium occur naturally and several more are man-made or are decay products of man-made isotopes. Radium is present in soil, minerals, foodstuffs, groundwater, and many common materials, including many used in construction. In communities where wells are used, drinking water can be an important source of ingested radium. Radium has been used commercially in luminous paints for watch and instrument dials and for other luminized objects. It has also been used for internal radiation therapy.

The primary sources of information on the health effects and dosimetry of radium isotopes come from extensive studies of ²²⁴Ra, ²²⁶Ra, and ²²⁸Ra in humans and experimental animals. These studies were motivated by the discovery of cancer and other debilitating effects associated with internal exposure to ²²⁶Ra and ²²⁸Ra. Later, similar effects were also found to be associated with internal exposure to ²²⁴Ra. The purpose of this chapter is to review the information on cancer induced by these three isotopes in humans and estimate the risks associated with their internal deposition.

All members of the world's population are presumably at risk, because each absorbs radium from food and water; as a working hypothesis, radiation is assumed to be carcinogenic even at the lowest dose levels, although there is no unequivocal evidence to support this hypothesis. Before concern developed over environmental exposure, attention was devoted primarily to exposure in the workplace, where the potential exists for the accidental uptake of radium at levels known to be harmful to a significant fraction of exposed individuals. As the practical concerns of radiation protection have shifted and knowledge has accumulated, there has been an evolution in the design and objectives of experimental animal studies and in the methods of collection, analysis, and presentation of human health effects data.

The first widespread effort to control accidental radium exposure was the abandonment of the technique of using the mouth to tip the paint-laden brushes used for application of luminous material containing ²²⁶Ra and sometimes ²²⁸Ra to the often small numerals on watch dials. This change occurred in 1925–1926 following reports and intensive discussion of short-term health effects such as ''radium jaw" in some dial painters. Shortly thereafter, experimental animal studies and the analysis of case reports on human effects focused on the determination of tolerance doses and radiation protection guides for the control of workplace exposure. These limits on radium intake or body content were designed to reduce the incidence of the then-known health effects to a level of insignificance. The question remained open, however, whether the health effects were threshold phenomena that would not occur below certain exposure or dose levels, or whether the risk would continue at some nonzero level until the exposure was removed altogether. The issue remains unresolved, but as a matter of philosophy, it is now commonly assumed that the so-called stochastic effects, cancer and genetic effects, are nonthreshold phenomena and that the so-called nonstochastic effects are threshold phenomena. Practical limitations imposed by statistical variation in the outcome of experiments make the threshold-nonthreshold issue for cancer essentially unresolvable by scientific study. For nonstochastic effects, apparent threshold doses vary with health endpoint. Low-level endpoints have not been examined with the same thoroughness as cancer. There is evidence that ^226,228Ra effects on bone occur at the histological level for doses near the limit of detectability. Whether these effects magnify other skeletal problems is unknown, but issues such as these leave the threshold-nonthreshold question open to further investigation.

Current efforts focus on the determination of risk, as a function of time and exposure, with emphasis on the low exposure levels where there is the greatest quantitative uncertainty. The presentation and analysis of quantitative data vary from study to study, making precise intercomparisons difficult. Occasionally, data from several studies have been analyzed by the same method, and this has helped to illuminate similarities and differences in response among ²²⁴Ra, ²²⁶Ra, and ²²⁸Ra.

Human health studies have grown from a case report phase into epidemiological studies devoted to the discovery of all significant health endpoints, with an emphasis on cancer but always with the recognition that other endpoints might also be significant. This chapter focuses on bone cancer and cancer of the paranasal sinuses and mastoid air cells because these effects are known to be associated with ²²⁴Ra or ^226,228Ra and are thought to be nonthreshold phenomena.

Several general sources of information exist on radium and its health effects, including portions of the reports from the United Nations Scientific Committee on the Effects of Atomic Radiation; The Effects of Irradiation on the Skeleton by Janet Vaughan; The Radiobiology of Radium and Thorotrast, edited by W. Gössner; The Delayed Effects of Bone Seeking Radionuclides, edited by C. W. Mays et al.; Volume 35, Issue 1, of Health Physics; the Supplement to Volume 44 of Health Physics; and publications of the Center for Human Radiobiology at Argonne National Laboratory, the Radioactivity Center at the Massachusetts Institute of Technology, the New Jersey Radium Research Project, the Radiobiology Laboratory at the University of California, Davis, and the Radiobiology Division at the University of Utah.

Chemistry and Physics of Radium

When injected into humans for therapeutic purposes or into experimental animals, radium is normally in the form of a solution of radium chloride or some other readily soluble ionic compound. Little research on the chemical form of radium in body fluids appears to have been conducted. The radium might exist in ionic form, although it is known to form complexes with some compounds of biological interest under appropriate physiological conditions; it apparently does not form complexes with amino acids.

Each isotope of radium gives rise to a series of radioactive daughter products that leads to a stable isotope of lead (Figure 4-1a and 4-1b). In addition to the primary radiation—alpha, beta, or both—indicated in the figures, most isotopes emit other radiation such as x rays, gamma rays, internal conversion electrons, and Auger electrons. In the analysis of radiation-effects data, the alpha particles emitted are considered to be the root cause of damage. This is because of the high linear energy transfer (LET) associated with alpha particles, compared with beta particles or other radiation, and the greater effectiveness of high-LET radiations in inducing cancer and various other endpoints, including killing, transformation, and mutation of cells.

Figure 4-1

a. Decay series for radium-228, a beta-particle emitter, and radium-224, an alpha-particle emitter, showing the principal isotopes present, the primary radiations emitted (α, β , or both), and the half-lives (s = second, m = minute, h (more...)

The decay products of radium, except radon, are atoms of solid materials. Radon is gaseous at room temperature and is not chemically reactive to any important degree. Unless physically trapped in a matrix, radon diffuses rapidly from its site of production. For ²²²Rn (whose half-life is very long compared with the time required for untrapped atoms within the body to diffuse into the blood supply), this rapid diffusion results in a major reduction of the radiation dose to tissues.

Retention and Distribution

Following entry into the circulatory system from the gut or lungs, radium is quickly distributed to body tissues, and a rapid decrease in its content in blood occurs. It later appears in the urine and feces, with the majority of excretion occurring by the fecal route. Retention in tissues decreases with time following attainment of maximal uptake not long after intake to blood. The loss is more rapid from soft than hard tissues, so there is a gradual shift in the distribution of body radium toward hard tissue, and ultimately, bone becomes the principal repository for radium in the body. The fundamental reason for this is the chemical similarity between calcium and radium. Because of its preference for bone, radium is commonly referred to as a bone seeker.

Various radiation effects have been attributed to radium, but the only noncontroversial ones are those associated with the deposition of radium in hard tissues. Two compartments are usually identified in the skeleton, a bone surface compartment in which the radium is retained for short periods and a bone volume compartment in which it is retained for long periods. A third compartment, which is not a repository for radium itself but which is relevant to the induction of health effects, consists of the pneumatized portions of the skull bones, that is, the paranasal sinuses and the air cells of the temporal bone (primarily the mastoid air cells), where radon and its progeny, the gaseous decay products of radium, accumulate.

Direct observation in vivo of retention in these three compartments is not possible, and what has been learned about them has been inferred from postmortem observations and modeling studies. During life, four quantities that can be monitored include whole-body content of radium, blood concentration, urinary excretion rate, and fecal excretion rate. These are supplemented by postmortem measurements of skeletal and soft-tissue content, observations of radium distribution within bone on a microscale, and measurements of radon gas content in the mastoid air cells.

For humans and some species of animals, an abundance of data is available on some of the observable quantities, but in no case have all the necessary data been collected. In general, the data from humans suffice to establish radium retention in the bone volume compartment. Animal data supplemented by models are required to estimate retention in the human bone surface, and human data combined with models of gas accumulation are applied to the pneumatized space compartment.

Figure 4-2 is a summary of data on the whole-body retention of radium in humans.²⁹ Whole-body retention diminishes as a power function of time. This observation has also been made for the retention of radium and other alkaline earths in animals Marshall and Onkelix³⁹ explained this retention in terms of the diffusion characteristics of alkaline earths in the skeleton.

Figure 4-2

Whole-body radium retention in humans. Summary of virtually all available data for adult man. The heavy curve represents the new model. Most of the points lie above the model curve for the first 1–2 days because no correction for fecal delay has (more...)

The excretion rate of radium can be determined by direct mea measurement in urine and feces or by determining the rate of change in whole-body retention with time. When radium levels in urine and feces are measured, by far the largest amount is found in the feces. In people with radium burdens of many years' duration, only 2% of the excreted radium exits through the kidneys. The other 98% passes out through the bowel.

At high radiation doses, whole-body retention is dose dependent. This observation was originally made on animals given high doses where retention, at a given time after injection, was found to increase with injection level. The most likely explanation is that tissue damage to the skeleton, at high doses, alters the retention pattern, primarily through the reduction in skeletal blood flow that results from the death of capillaries and other small vessels and through the inhibition of bone remodeling, a process known to be important for the release of radium from bone. A recent examination of data on whole-body radium retention in humans revealed that the excretion rate diminished with increasing body burden.⁷⁰ Absolute retention could not be studied, because the initial intake was unknown, but the data imply the existence of a dose-dependent retention similar to that observed in animals. Subnormal excretion rate can be linked with the apparent subnormal remodeling rates in high-dose radium cases.⁷⁷

Radium has an affinity for hard tissue because of its chemical similarity to calcium. It does, however, deposit in soft tissue and there is a potential for radiation effects in these tissues. The data on human soft-tissue retention were recently reviewed.⁷⁴ The rate of release from soft tissue exceeds that for the body as a whole, which is another way of stating that the proportion of total body radium that eventually resides in the skeleton increases with time.

Postmortem skeletal retention has been studied in animals and in the remains of a few humans with known injection levels. Otherwise, the retention in bone is estimated by models.

Autoradiographic studies³⁷ of alkaline earth uptake by bone soon after the alkaline earth was injected into animals revealed the existence of two distinct compartments in bone (see Figure 4-3), a short-term compartment associated with surface deposition, and a long-term compartment associated with volume deposition. The take and release of activity into and out of the surface compartment was studied quantitatively in animals and was found to be closely related to the time dependence of activity in the blood.⁶⁵ Mathematical analysis of the relationship showed that bone surfaces behaved as a single compartment in constant exchange with the blood.³⁷ This model for the kinetics of bone surface retention in animals was adopted for man and integrated into the ICRP model for alkaline earth metabolism, in which it became the basis for distinguishing between retention in bone volume and at bone surfaces. This is an instance in which an extrapolation of animal data to humans has played an important role.

Figure 4-3

Autoradiograph of bone from the distal left femur of a former radium-dial painter showing hotspots (black areas) and diffuse radioactivity (gray areas).

A mechanistic model for alkaline earth metabolism²⁹ was developed by the ICRP to describe the retention of calcium, strontium, barium, and radium in the human body and in human soft tissue, bone volume, bone surfaces, and blood. Separate retention functions are given for each of these compartments. When the model is used for radium, careful attention should be paid to the constraints placed on the model by data on radium retention in human soft tissues.⁷⁴ Because of the mathematical complexity of the retention functions, some investigators have fitted simpler functions to the ICRP model.

These simpler functions have no mechanistic interpretation, but they do make some calculations easier.

The kinetics of radon accumulation in the pneumatized air spaces are determined by the kinetics of radium in the surrounding bone, the rate of diffusion from bone through the intervening tissue to the air cavity, and the rate of clearance through the ventilatory ducts and the circulatory system. Diffusion models for the sinuses have not been proposed, but work has been done on the movement of ²²⁰Rn through tissue adjacent to bone surfaces. Clearance through the ventilatory ducts is rapid when they are open. The eustachian tube provides ventilation for the middle ear and pneumatized portions of the temporal bone. This duct is normally closed, and clearance By this pathway is negligible. The sinus ducts are normally open but can Be plugged by mucus or the swelling of mucosal tissues during illness. When these ducts are open, clearance is almost exclusively through them. Clearance half-times for the frontal and maxillary sinuses are a few minutes when the ducts are open. Otherwise, clearance half-times are about 100 rain and are determined by the blood flow through mucosal tissues.⁷³ The radioactive half-lives of the radon isotopes—55 s for ²²⁰Rn and 3.8 days for ²²²Rn—are quite different from their clearance half-times. In effect, essentially all the ²²⁰ Rn that diffuses into the pneumatized air space decays there Before it can be cleared, but essentially all the ²²²Rn that reaches the pneumatized air space is cleared before it can decay. These relationships have important dosimetric implications.

Bone Cancer

Frequency and Cell Type

Radium deposited in bone irradiates the cells of that tissue, eventually causing sarcomas in a large fraction of subjects exposed to high doses. The first case of bone sarcoma associated with ^226,228Ra exposure was a tumor of the scapula reported in 1929, 2 yr after diagnosis in a woman who had earlier worked as a radium-dial painter.⁴² Bone tumors among children injected with ²²⁴Ra for therapeutic purposes were reported in 1962 among persons treated between 1946 and 1951.⁸⁷

Spontaneously occurring bone tumors are rare. Sarcomas of the bones and joints comprise only 0.24% of microscopically confirmed malignancies reported by the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program.⁵² The chance of contracting bone sarcoma during a lifetime is less than 0.1%.

Some 87 bone sarcomas have occurred in 85 persons exposed to ^226,228 Ra among the 4,775 persons for whom there has been at least one determination of vital status. Multiple sarcomas not confirmed as either primary or secondary are suspected or known to have occurred in several other subjects. A total of 66 sarcomas have occurred in 64 subjects among 2,403 subjects for whom there is an estimate of skeletal dose; fewer than 2 sarcomas would be expected. Many of the 2,403 subjects are still alive. Tumor frequencies for axial and appendicular skeleton are shown in Table 4-1. The frequencies for different bone groups are axial skeleton-skull (3), mandible (1), ribs (2), sternebrae (1), vertebrae (1), appendicular skeleton-scapulae (2), humeri (6), radii (2), ulnae (1), pelvis (10), femora (22), tibiae (7), fibulae (1), legs (2; bones unspecified), feet and hands (5; bones unspecified).

TABLE 4-1

Locations of Bone Sarcomas among Persons Exposed to ²²⁴ Ra and ^226,228Ra for Whom Skeletal Dose Estimates Are Available.

Some 55 sarcomas of bone have occurred in 53 of 898 ²²⁴Ra-exposed patients whose health status is evaluated triennially.⁴⁶ Two primary sarcomas occurred in 2 subjects. Locations are shown in Table 4-1 for 49 tumors among 47 subjects for whom there is an estimate of skeletal dose.

In Table 4-1 note the low tumor yield of the axial compared with the appendicular skeleton. In an earlier summary for 24 ²²⁴Ra-induced osteosarcomas,⁹⁰ 21% occurred in the axial skeleton. These percentages contrast sharply with the results for beagles injected with ²²⁶Ra, in which osteosarcomas were about equally divided between the axial and appendicular skeletons and one-quarter of the tumors appeared in the vertebrae.⁹⁰

Histologic type has been confirmed by microscopic examination of 45 tumors from 44 persons exposed to ^226,228Ra for whom dose estimates are available; there were 27 osteosarcomas, 16 fibrosarcomas, 1 spindle cell sarcoma, and 1 pleomorphic sarcoma. The distributions of histologic types for the 47 subjects exposed to ²²⁴Ra with bone sarcoma and a skeletal dose estimate are 39 osteosarcomas, 1 fibrosarcoma, 1 pleomorphic sarcoma, 4 chondrosarcomas, 1 osteolytic sarcoma, and 3 bone sarcomas of unspecified type. The distribution of tumor types is not likely to undergo major changes in the future; the group of ^226,228Ra-exposed patients at high risk is dwindling due to the natural mortality of old age and the rate of tumor appearance among ²²⁴Ra-exposed patients has dropped to zero in recent years.⁴⁶

The distribution of histologic types for radium-induced tumors is compared in Table 4-2 with that reported for naturally occurring bone tumors.¹¹ The data have been divided into two groups according to age of record for the tumor. In some cases, this is the age at death and in others this is the age at which the presence of the tumor can be definitely established from the information available. The data have been normalized to the frequency for osteosarcoma and limited to the three principal radiogenic types: osteosarcoma, chondrosarcoma, and fibrosarcoma.

TABLE 4-2

Relative Frequencies for Radium-Induced and Naturally Occurring Tumors by Age Group.

Under age 30, the relative frequencies for radiogenic tumors are about the same as those for naturally occurring tumors. The total numbers of tumors available are too small to assign significance to the small differences in relative frequencies for a given histologic type. Over age 30, the situation is different. Distinctly lower relative frequencies occur for chondrosarcoma and fibrosarcoma induced by ²²⁴Ra compared with these same types that occur spontaneously. The relative frequencies for fibrosarcomas induced by ²²⁴Ra and ^226,228 Ra are also different, as are the relative frequencies for chondrosarcomas induced by ^226,228Ra and naturally occurring chondrosarcomas. Thus, the spectrum of tumor types appears to be shifted from the naturally occurring spectrum when the tumors are induced by radium.

Radium-226 and Radium-228 Bone Cancer

The original cases of radium poisoning were discovered by symptom, not by random selection from a defined population. This method of selection, therefore, made such cases of questionable suitability for inclusion in data analyses designed to determine the probability of tumor induction in an unbiased fashion. To circumvent this problem, two strategies have been developed: (1) classification of the cases according to their epidemiological suitability, on a scale of 1 to 5, with 5 representing the least suitable and therefore the most likely to cause bias and 1 representing the most suitable and therefore the least likely to cause bias; and (2) definition of subgroups of the whole population according to objective criteria presumably unrelated to tumor risk, for example, by year of first exposure and type of exposure. The latter method does not, in effect, correct for selection bias because there is no way to select against such cases. For radium-dial painters, however, the number of persons estimated to have worked in the industry is not too much greater than the number of subjects that have been located and identified by name.⁶⁷ This fact implies that coverage of the radium-dial painter segment of the population is reasonably good, thus reducing concerns over selection bias.

The first comprehensive graphical presentations of the dose-response data were made by Evans.¹⁵ In that study both tumor types (bone sarcoma and head carcinoma) were lumped together, and the incidence data were expressed as the number of persons with tumor divided by the total number known to have received the same range band of skeletal radiation dose. These were plotted against a variety of dose variables, including absorbed dose to the skeleton from ²²⁶Ra and ²²⁸Ra, pure radium equivalent, and time-weighted absorbed dose, referred to as cumulative rad years. This type of analysis was used by Evans¹⁵ in several publications, some of which employed epidemiological suitability classifications to control for case selection bias. Regardless of the dose variable used, the scatter diagram indicated a nonlinear dose-response relationship, a qualitative judgment that was substantiated by chi-squared tests of the linear functional form against the data.

Concern over the shape of the dose-response relationship has been a dominant theme in the analyses and discussions of the data related to human exposure to radium. In simple terms, the main issue has been linear or nonlinear, threshold or nonthreshold. Evans et al. provided an interesting and informative commentary on the background and misapplications of the linear nonthreshold hypothesis.¹⁷

Concurrently, Mays and Lloyd⁴⁴ analyzed the data on bone tumor induction by using Evans' measures of tumor incidence and dosage without correction for selection bias and presented the results in a graphic form that leaves a strong visual impression of linearity, but which, when subjected to statistical analysis, is shown to be nonlinear with high probability. Although the conclusions to be drawn from Evans' and Mays' analyses are the same—that a linear nonthreshold analysis of the data significantly overpredicts the observed tumor incidence at low doses—there is a striking difference in the appearance of the data plots, as shown in Figure 4-4, in which the results of studies by the two authors are presented side by side.

Figure 4-4

Dose-response relationships of Evans et al. (a), Mays and Lloyd (b), and Rowland et al. (c).

Following consolidation of U.S. radium research at a single center in October 1969, the data from both studies were combined and analyzed in a series of papers by Rowland and colleagues.^66–69 Bone tumors and carcinomas of the paranasal sinuses and mastoid air cells were dealt with separately, epidemiological suitability classifications were dropped, incidence was redefined to account for years at risk, and dose was usually quantified in terms of a weighted sum of the total systemic intakes of ²²⁶Ra and ²²⁸Ra, although there were analyses in which mean skeletal dose was used. The use of intake as the dose parameter rested on the fact that it is a time-independent quantity whose value for each individual subject remains constant as a population ages. In contrast, mean skeletal dose changes with time, causing a gradual shift of cases between dose bands and confusing the intercomparison of data analyses carried out over a period of years. The outcome of the analyses of Rowland and colleagues was the same whether intake or average skeletal dose was employed, and for comparison with the work of Evans and Mays and their coworkers, analyses based on average skeletal dose will be used for illustration. Another difference between the analyses done by Rowland et al. and those done earlier was division of the radium-exposed subjects into subpopulations defined by type of exposure, that is, radium-dial workers (mostly dial painters), those medically exposed, and others. In this way, some problems of selection bias could be avoided, because most radium-dial workers were identified by search, and coverage of the radium-dial worker groups was considered to be high. Coverage of other groups, especially those with medical exposure, was considered low, and many subjects were selected by symptom. Dose-response data were fitted by a linear-quadratic-exponential expression:

where D is estimated systemic intake. Equation 4-1 was modified from the general form adopted in the BEIR III report:⁵⁴

As with Evans et al.'s work,¹⁷ the data were plotted against the logarithm of dose so that the low-dose region was not obscured.

A plot of the bone sarcoma data for a population subgroup defined as female radium-dial workers first exposed before 1930 is shown in Figure 4-4. The dissimilarities, primarily between the plots of Evans et al. and Rowland et al., are from the use of person-years at risk in the definition of tumor incidence, from the inclusion of both groups of radium-induced tumor, and the use of different weighting factors in the summation of ²²⁶Ra and ²²⁸Ra dose. In the Evans et al. analysis, ²²⁶Ra and ²²⁸Ra dose contributions were weighted equally; in Rowland et al.'s analysis, the ²²⁸Ra dose was given a weight 1.5 times that of ²²⁶Ra.

The functional form in the analysis of Rowland et al. that provided the best fit to the data as judged by the chi-squared test, was (C + β D²) exp(-γD), although three other forms provided acceptable fits: C + αD + β D², (C + αD) exp(-γD), and (C + αD + β D²) exp(-γ D). If forms with negative coefficients are eliminated, as postulated by the model, then only (C + αD) exp(-γD) from this latter group provided an acceptable fit, but it had a chi-squared probability (0.06) close to the rejection level (0.05). Forms with positive coefficients, which were rejected on the basis of goodness of fit, were C + αD and C + β D². Rowland et al. concluded that linear dose-response function was incapable of describing the data over the full range of doses.

Three other analyses of the data relevant to the shape of the dose-response curve are noteworthy. The first is that of Rowland et al.⁶⁷ in which estimated systemic intake (D) rather than average skeletal absorbed dose was used as the dose parameter and functions of the form (C + αD + β D²) exp(-γD) were fitted to the data. The findings were similar to those described above. For female radium-dial workers first employed before 1930, the only acceptable fit to the data on bone sarcomas per person-year at risk was provided by the functional form (C + β D²) exp(-γD), which was obtained from the more general expression by setting α = 0. When the population was later broadened to include all female radium-dial workers first employed before 1950⁶⁹ for whom there was an estimate of radium exposure based on measurement of body radioactivity, a much larger group than female radium-dial workers first employed before 1930 (1,468 versus 759), the only acceptable fit was again provided by the functional form (C + β D²) exp(-γD). When the size of the study group was reduced by changing the criterion for acceptance into the group from year of first entry into the industry to year of first measurement of body radioactivity while living, the observed number of bone tumors dropped from 42 to 13, because radioactivity in many persons was first measured after death. Under these circumstances, the forms C + αD and (C + β D²) exp(-γD) gave acceptable fits.

The second analysis is that of Marshall and Groer,³⁸ in which a carefully constructed theoretical model was fitted to bone-cancer incidence data. The model was based on a series of three differential equations that described the dynamics of cell survival, replacement, and transformation when bone is irradiated by alpha particles. The outcome of the fitting procedure was presented in graphic form, with total unweighted estimated systemic intake of ²²⁶Ra and ²²⁸Ra normalized to body weight as the dose parameter. Cumulative incidence, which is the total number of tumors per intake group divided by the numbers of persons alive in that group at the start of observation, was the response parameter. An acceptable fit, as judged by a chi-squared criterion, was obtained. At low doses, the model predicts a tumor rate (probability of observing a tumor per unit time) that is proportional to the square of endosteal bone tissue absorbed dose. In the model, this dose is directly proportional to the average skeletal dose, and tumor rate is an analog of the response parameter, which is bone sarcomas per person-year at risk. Thus, the model and the Rowland et al. analysis are closely parallel and, as might be expected, lead to the same general conclusion that the response at low doses [where exp(-γD) ≅ 1] is best described by a function that varies with the square of the absorbed dose. The analysis of Marshall and Groer³⁸ is noteworthy, not only because it provides a good fit to the data but also because it links dose and events at the cellular level to epidemiological data, an essential step if the results of experimental research at the cellular level are to play a serious role in the estimation of tumor risk at low doses.

The third analysis was carried out by Raabe et. al.,^61,62 with time to death by bone cancer and average skeletal dose rate as the response and dose parameters, respectively. The analysis is most relevant to the question of practical threshold and will be discussed again in that context. Raabe et al. employed a log-normal dose-rate, time-response model that was fitted to the data and that could be used to determine bone-cancer incidence, measured as a percentage of those at risk, versus absorbed skeletal radiation dose. When plotted, the model shows a nonlinear dose-response relationship for any given time after exposure.

Evans, Mays, and Rowland and their colleagues presented explicit numerical values or functions based on their fits to the radium tumor data. For Evans' analysis, the percent tumor cumulative incidence for bone sarcomas plus head carcinomas is constant at 28 ± 6% for mean skeletal doses between 1,000 and 50,000 rad. No fitted value is given for doses below 1,000 rad, but all data points in this range are at zero incidence. Error bars on the points vary in size, and are all less than about 6% cumulative incidence (Figure 4-4). It is clear, therefore, that a nonzero function could be fitted to these data but would have numerical values substantially less than 28%.

For the percent of exposed persons with bone sarcomas, Mays and Lloyd⁴⁴ give 0.0046% D _s, where D _s is the sum of the average skeletal doses for ²²⁶Ra and ²²⁸Ra, in rad.

In the analysis by Rowland et al. ^67,68 based on dose, equations that give an acceptable fit are:

where the risk coefficient I equals the number of bone sarcomas per person-year at risk that begin to appear after a 5 yr latent period, and D _s is the average skeletal dose from ²²⁶Ra plus 1.5 times the average skeletal dose from ²²⁸Ra, expressed in rad. For the analyses based on intake, the equation that gives an acceptable fit is:

where I is bone sarcomas per person-year at risk, and D _i is the total systemic intake of ²²⁶Ra plus 2.5 times the total systemic intake of ²²⁸Ra, expressed in microcuries. In the latter analysis,⁶⁹ the only acceptable fit based on year of entry into the study is:

where I and D _i are as defined above. The equations based on year of first measurement of body radioactivity are:

With attention now focused on exposure levels well below those at which tumors have been observed, it is natural to exploit functions such as those presented above for radiogenic risk estimation. The radiogenic risk equals the total risk given by one of the preceding expressions minus the natural tumor risk. This may lead to negative values at low exposures. For example, if D _s = 0.5 rad, which is approximately equal to the lifetime skeletal dose associated with the intake of 2 liters/day of water containing the Environmental Protection Agency's maximum concentration limit of 5 pCi/liter, the expression of Mays and Lloyd⁴⁴ would predict a total risk of 0.0023%. With a lifetime natural tumor risk of 0.1%, the radiogenic risk would be -0.0977%. Comparable examples can be given for each expression of Rowland et al. that contains an exponential factor. Such negative values follow logically from the mathematical models used to fit the data and underscore the inaccuracy and uncertainty associated with evaluating the risk far below the range of exposures at which tumors have been observed.

Negative values have been avoided in practical applications by redefining the dose-response functions at low exposure levels. For the Mays and Lloyd⁴⁴ function, this consists of setting the radiogenic risk equal to the total risk rather than to the total risk minus the natural risk. For the functions of Rowland et al. that contain an exponential factor, the natural tumor rate is set equal to zero, and the resulting expression is then defined as the radiogenic risk. For example, when the risk coefficient is:

the radiogenic risk would be:

For functions that lack an exponential factor, such as I = 1.75 × 10^-5 + (2.0 ± 0.6) × 10^-5 D _i, redefinition is not required to avoid negative expected values, and radiogenic risk is set equal to the difference between total risk and natural risk.

There have been two systematic investigations of the ^226,228Ra data related to the uncertainty in risk at low doses. Rowland et al.⁶⁹ examined the class of functions I = (C + αD _i + βD_i ²) exp(-γD _i) with positive coefficients, not all of which were determined by least-square fitting to the data, based on year of entry and found that:

determined the upper and lower boundaries (I _u and I _l, respectively) of an envelope of curves that provided acceptable fits to the data, as judged by a chi-squared criterion. When the radiogenic risk functions (I _u - 0.7 × 10^-5) and (I _l - 0.7 × 10^-5) are used to determine a range of values based on the envelope boundaries, a measure of the uncertainty in estimated bone sarcoma risk at low doses can be formed as:

where I is the best-fit function [0.7 × 10^-5 + 7.0 × 10^-8 D _i ²]exp(-1.1 × 10^-3 D _i), based on year of entry. This ratio increases monotonically with decreasing intake, from a value of 1.5 at D _i = 100 µCi to a value of 480 at D _i = 0.5 µCi.

Schlenker⁷⁴ presented a series of analyses of the ^226,228Ra tumor data in the low range of intakes at which no tumors were observed but to which substantial numbers of subjects were exposed. For each of the seven intake groupings in this range (e.g., 0.5–1, 1–2.5, 2.5–5), there was about a 5% chance that the true tumor rate exceeded 10^-3 bone sarcomas per person-year when no tumors were observed, and there was a 48% chance that the true tumor rate, summed over all seven intake groups exceeded the rate predicted by the best-fit function I = (10^-5 + 6.8 × 10^-8 D _i ²)exp(-1.1 × 10^-3 D _i). With smooth curves, this analysis defined envelopes for which there was a 9, 68, or 95% chance that the true tumor rate summed over the seven intake groups fell between the envelope boundaries when no tumors were observed. The 9% envelope was obtained by allowing the parameters in the function to vary by 2 standard errors on either side of the mean and emphasizes that the standard errors obtained by least-square fitting underestimate the uncertainty at low doses. Figure 4-5 shows the results of this analysis, and Table 4-3 gives the equations for the envelope boundaries.

Figure 4-5

Dose-response envelopes for ^226,228Ra.

TABLE 4-3

Equations for the Functions I _u and I _l That Define the Dose-Response Envelopes in Figure 4-5.

This work allows one to specify a central value for the risk, based on the best-fit function and a confidence range based on the envelopes. For example, the central value of total risk, including that from natural causes, is I = (10^-5 + 6.8 × 10^-8 D _i ²)exp(-1.1 × 10^-3 D _i) with 95% confidence that total risk lies between I _l = 10^-5 and I _u = 10^-5 + 1.6 × 10^-5 D _i - 3.6 × 10^-8 D _i ² for D _i between 0.5 and 100 µCi. The ratio of the 95% confidence interval range, for radiogenic risk, to the central value,

increases with decreasing intake from 1.7 at D _i = 100 µCi to 700 at D _i = 0.5 µCi, the lower boundary of the lowest intake cohort used when fitting functions to the data.

When radiogenic risk is determined by setting the natural tumor rate equal to 0 in the expressions for total risk and by eliminating the natural tumor rate (10^-5/yr) from the denominator in Equation 4-14, the value of the ratio increases more slowly, reaching 470 at D _i = 0.5 µCi. At D _i = 0.05 µCi, the total systemic intake in 70 yr for a person drinking 2 liters of water per day at the Environmental Protection Agency's maximum contaminant level of 5 pCi/liter, the ratio is 4,700. These high ratios emphasize, in quantitative terms, our ignorance of risk at low exposure levels.

The risk envelopes defined by these analyses are not unique. Other functions can be determined that meet this 95% probability criterion. This emphasizes that there is no unique way to specify the uncertainty in risk at low exposures when the shape of the dose-response curve is unknown. Regardless of the functions selected as envelope boundaries, however, the percent uncertainty in the risk cannot be materially reduced.

²²⁴Ra Bone Cancer

Internal radiation therapy has been used in Europe for more than 40 yr for the treatment of various diseases. Between 1944 and 1951 it was injected in the form of Peteosthor, a preparation containing ²²⁴Ra, eosin, and colloidal platinum, primarily for the treatment of tuberculosis and ankylosing spondylitis. Its use with children came to an end in 1951, following the realization that growth retardation could result and that it was ineffective in the treatment of tuberculosis. Since then it has been used with adults as a clinically successful treatment for the debilitating pain of ankylosing spondylitis. Platinum and eosin, once thought to focus the uptake of ²²⁴Ra at sites of disease development, have been proven ineffective and are no longer used.

Two extensive studies of the adverse health effects of ²²⁴Ra are under way in Germany. Roughly 900 persons who were treated with Peteosthor as children or adults during the period 1946–1951 have been followed by Spiess and colleagues^84–86 for more than 30 yr and have shown a variety of effects, the best known of which is bone cancer. To supplement these investigations of high-level exposure, a second study was initiated in 1971 and now includes more than 1,400 individuals treated with small doses of ²²⁴Ra for ankylosing spondylitis and more than 1,500 additional patients with ankylosing spondylitis treated with other forms of therapy who serve as controls.

As with other studies, the shape of the dose-response curve is an important issue. Based on their treatment of the data, Mays et al.⁴⁹ made the following observation: ''We have fit a variety of dose-response relationships through our follow-up data, including linear (y = ax), linear multiplied by a protraction factor, dose-squared exponential (y = ax ² e ^-kx), and a threshold function. None can be rejected because of the scatter in our human data." Rowland⁶⁴ published linear and dose-squared exponential relationships that provided good visual fits to the data. Recent analyses with a proportional hazards model led to a modification of the statement about the adequacy of the linear curve, as will be discussed later. However, the change was not so great as to alter the basic conclusion that the data have too little statistical strength to distinguish between various mathematical expressions for the dose-response curve. As a convenient working hypothesis, in several papers it has been assumed that the linear form is the correct one, leading to analyses that are illuminating and easily understood.

In the first dose-response analyses, average skeletal dose was adopted as the dose parameter, and details of the dose calculations were presented. With only two exceptions, average skeletal dose computed in the manner described at that time has been used as the dose parameter in all subsequent analyses. As a response parameter, the number of bone sarcomas that have appeared divided by the number of persons known to have been exposed within a dose group was used. The data for persons exposed as juveniles (less than 21 yr of age) were analyzed separately from the data for persons exposed as adults, and different linear dose-response functions that fit the data adequately over the full range of doses were obtained.⁸⁵ The linear slope for juveniles, 1.4%/100 rad, was twice that for adults, 0.7%/100 rad. The analysis took into account tumors appearing between 14 and 21 yr after the start of exposure in 43 subjects that received a known dose. These constitute about 85% of the subjects with bone sarcoma on which the most recent analyses have been based. The importance of this work lies in the fact that it shows the maximum difference in radiosensitivity between juvenile and adult exposures for this study. In later work, juvenile-adult differences have not been reported.

The removal of the difference came in two steps associated with analyses of the influence of dose protraction on tumor induction. Based on a suggestion by Muller drawn from his observations of mice, Speiss and Mays⁸⁶ reanalyzed their ²²⁴Ra data in an effort to determine whether there was an association between dose protraction and tumor yield. In the analyses, a linear dose-response relationship was postulated, and the data were sorted according to the time period over which ²²⁴Ra was administered. The found that the slope of the linear dose-response curve increased with increasing time period, suggesting that bone-cancer incidence increased with decreasing average skeletal dose rate, in accordance with results in mice. Although the change of tumor incidence with exposure duration was not statistically significant, an increase did occur both for juveniles and adults. In a subsequent analysis,⁴⁶ the data on juveniles and adults were merged, and an additional tumor was included for adults, bringing the number of subjects with tumors and known dose to 48. A single function was fitted to these data to describe the change of the dose-response curve slope with the length of time over which injections were given:

where y is the number of bone sarcomas per million person-rad and x is the length of the injection span, in months. The asymptotic value of this function is 200 bone sarcomas/million person-rad, which is considered applicable both to childhood and adult exposure. For comparison with the values given previously for juveniles and adults separately, this is 2.0% incidence per 100 rad, which is somewhat higher than either of the previous values.

The case for a dose rate or dose-protraction effect rests on the observation of an association of the linear dose-response slope with dose rate in humans and the unequivocal appearance of a dose-protraction effect in mice and rats. Though one might wish to dispute its existence in humans on statistical grounds in order to defend a claim for greater childhood radiosensitivity, it would seem uneconomical to do so until there is clear evidence of greater radiosensitivity to alpha radiation for the induction of bone cancer in the young of another species.

The first analysis to take account of competing risks and loss to followup⁷⁴ was based on a life-table analysis of data collected⁸⁸ for persons 16 yr of age and older. Cumulative incidence, computed as the product of survival probabilities in the life table,¹⁰ was used as the measure of response with errors based on approximations by Stehney. As the dose parameter, absorbed dose in endosteal tissue was used, computed from the injection levels, in micrograms per kilogram, using conversion factors based on body weight and relative distribution factors similar to those of Marshall et al.⁴⁰ but altered to take into account the dependence of stopping power on energy. The functional form found to provide a best fit to the data was:

where ν/N is the cumulative incidence, and D _e is the endosteal dose.

Not long afterward, Mays and Spiess⁴⁵ published a life-table analysis in which cumulative incidence was computed annually from the date of first injection by summing annual tumor occurrence probabilities. For each year, the cumulative incidence so obtained was divided by the average value of the mean skeletal dose for subjects within the group, in effect yielding the slope of a linear dose-response curve for the data. Adults and juveniles were treated separately. The resultant graph of dose-response curve slopes versus years of follow-up is shown in Figure 4-6. Although the points for adults always lie below those for juveniles, there is always substantial statistical overlap. In a subsequent life-table analysis, in which the same methods were used but 38 cases for whom there were not dose estimates were excluded, the points for juveniles and adults lie somewhat further apart. This type of analysis updates the one originally conducted for this group of subjects in which juvenile radiosensitivity was reported to be a factor of 2 higher than adult radiosensitivity. According to the latest life-table analysis, the risk to juveniles (188 ± 32 bone sarcomas/10⁶ person-rad) is 1.4 times the risk to adults (133 ± 36 bone sarcomas/10⁶ person-rad). Presumably, if dose protraction were taken into account by the life-table analysis, the difference between juveniles and adults would vanish.

Figure 4-6

The cumulative tumor risk (bone sarcomas/10⁶ person-rad) was similar in the juvenile and adult patients under the dosimetric assumptions used. The standard deviation for each point is shown. Source: Mays and Spiess.

The third analysis that corrects for competing risks was performed by Chemelevsky et al.⁹ using a proportional hazards model. The data for juveniles and adults was separated into different dose groups, a step not taken with the life-table analysis of Mays and Spiess.⁴⁵ This, in effect, frees the analysis from the assumption of a linear dose-response relationship, implicit in the Mays and Spiess analysis. Estimates of the cumulative tumor rate (incidence) versus time after first injection were obtained, and when those for juveniles and adults in comparable dose groups were compared, no difference in either the magnitude or the growth of cumulative tumor rate with time was found between the two age groups. The cumulative tumor rate for juveniles and adults at 25 yr after injection, a time after which, it is now thought, no more tumors will occur, were merged into a single data set and fitted with a linear-quadratic exponential relationship:

where R is the probability that a tumor will occur per person-gray and D _s is the average skeletal dose in gray (1 Gy is 100 rad). This curve and the data points are shown in Figure 4-7. The error bars on each point are a greater fraction of the value for the point here than in Figure 4-6, because the subdivision into dose groups has substantially reduced the number of subjects that contributes to each datum point. A linear function was fitted to the data over the full range of doses, but the fit was rejected by a statistical test for goodness of fit that yielded a P value of 0.02. This is the first report of an explicit test of linearity that has resulted in rejection. The best-fit function, however, does contain a linear term, in contrast to the best-fit functions for the data on ^226,228Ra. This means that when doses are low enough, the risk varies linearly with dose. The same observation can be made for the function 1 - exp(-0.00003D) for the probability of tumor induction developed from the life-table analysis of Schlenker.⁷⁴

Figure 4-7

Risk per person per gray versus mean skeletal dose. The points with their standard errors result from the proportional hazards analysis of Chemelevsky et al.

The data points in Figure 4-7 for juveniles and adults are not separable from one another, and the difference between juvenile and adult radiosensitivity has completely disappeared in this analysis. From this, we can conclude that much, and perhaps all, of the difference in radiosensitivity between juveniles and adults originally reported was due to the failure to take into account competing risks and loss to follow-up. If a dose-protraction effect were included in the analysis, there might be a reversal of the original situation, with adults having the greater radiosensitivity. The results of this series of studies of bone sarcoma incidence among ²²⁴Ra-exposed subjects extending over a period of 15 yr underscore the importance of repeated scrutiny of unique sets of data. Whether due to competing risks, dose protraction, or a combination, it is clear that differential radiosensitivity for this group of subjects is a hypothesis that cannot be supported.

As of December 1982, the average followup time was 16 yr for patients injected after 1951 with lower doses of ²²⁴Ra for the treatment of ankylosing spondylitis.⁹³ Of 1,426 patients who had been traced, the vital status for 1,095 of them was known. Of these, 363 died and three bone cancers, one fibrosarcoma, one reticulum cell sarcoma, and one multiple myeloma were recorded. The average dose for the exposed group, based on patients for whom there were extant records of treatment level, was 65 rad. In discussing these cases, Wick and Gössner⁹³ noted that three cases of bone cancer were within the range expected for naturally occurring tumors and also within the range expected from a linear extrapolation downward to lower doses from the Spiess et al.⁸⁸ series. However, 80% of the bone tumors in the this series, for which histologic type is known, are osteosarcomas, while fibrosarcomas and reticulum cell sarcomas each represent only about 2% of the total, and multiple myeloma was not observed at all. Based on this, the chance of randomly selecting three tumors from the this distribution and coming up with no osteosarcomas is about (0.2)³ = 0.008, throwing the weight of evidence in favor of a nonradiogenic origin for the three bone cancers found in this study.^93,94 However, this could occur if there were a dramatic change in the distribution of histologic types for tumors induced by ²²⁴Ra at doses below about 90 rad, which is approximately the lower limit for tumor induction in the Spiess et al.⁸⁸ series. If the tumors are nonradiogenic, then the linear extrapolation gives a substantial over prediction of the risk at low doses, just as a linear extrapolation of the ^226,228Ra data overpredict the risk from these isotopes at low doses.^17,44

Schlenker⁷⁴ has provided a confidence interval analysis of the Spiess et al.⁸⁸ data in the region of zero observed tumor incidence to parallel that for ^226,228Ra. The results are shown in Figure 4-8. There is a 14% probability that the expected number of tumors lies within the shaded region, defined by allowing the parameter value in Equation 4–16 to vary by 2 standard errors about the mean, and a 68% probability that it lies between the solid line that is nearly coincident with the upper boundary of the shaded region and the lower solid curve. The shaded region emphasizes that standard errors obtained by least-square fitting underestimate the uncertainty in risk at low doses. There is a 95% probability that the expected number lies between the dashed boundaries. As with ^226,228Ra, the curves in Figure 4-8 can be used to establish confidence limits for risk estimates at low doses, although it is to be understood that these limits are not unique, because the shape of the dose-response curve is unknown. As an example, the upper boundaries of the 95% confidence envelope for total cumulative incidence corrected for competing risks are:

Figure 4-8

Dose-response envelopes for ²²⁴Ra from equation 4–16. The upper curve of the 68% envelope is nearly coincident with the upper boundary of the shaded envelope.

Those for the lower boundary are:

The ratio of the 95% confidence interval range for radiogenic risk to the radiogenic risk defined by the central value function

where 3 × 10^-5 is the natural risk adapted here. This ratio increases monotonically with decreasing endosteal dose, from 1.8 at 500 rad to 220 at 25 rad, which is the lower boundary of the lowest dose cohort used in Schlenker's⁷⁴ analysis.

Carcinoma of the Paranasal Sinuses and Mastoid Air Cells

The paranasal sinuses are cavities in the cranial bones that exchange air and mucus with the nasal cavity through a small ostium. The sinuses are present as bilateral pairs and, in adulthood, have irregular shapes that may differ substantially in volume between the left and right sides. The ethmoid sinuses form several groups of interconnecting air cells, on either side of the midline, that vary in number and size between individuals.⁹² The sinus surfaces are lined with a mucous membrane that is contiguous with the nasal mucosa and consists of a connective tissue layer attached to bone along its lower margin and to a layer of epithelium along its upper margin. The cilia transport mucus in a more or less continuous sheet across the epithelial surface toward the ostium.¹³

The mastoid air cells, like the ethmoid sinuses, are groups of interconnecting air cavities located bilaterally in the left and right temporal bones. The mastoid air cells communicate with the nasopharynx through the middle ear and the eustachian tube. Radium-induced carcinomas in the temporal bone are always assigned to the mastoid air cells, but the petrous air cells cannot be logically excluded as a site of origin.

The mucosal lining of the mastoid air cells is thinner than the lining of the sinuses. The epithelium is of squamous or cuboidal type with scattered ciliated cells but no goblet cells. It shows no signs of significant secretory activity but is always moist. The typical adult maxillary cavity has a volume of about 13 cm³; one frontal sinus has a volume of about 4.0 cm³, and one sphenoid sinus has a volume of about 3.5 cm³. The collective volume of one set of ethmoid air cells is about 3.5 cm³; there are nine cells on the average,⁹² for an average volume per cell of 0.4 cm³. The pneumatized portion of one mastoid process has a volume of about 9.2 cm³. The individual cells range from 0.1 to more than 1 cm across and are too numerous to be counted. The distance across a typical air cell is 0.2 cm,⁷³ equivalent to a volume of about 0.004 cm³ if the cell were spherical.

The total thickness of the mucosa, based on the results of various investigators, ranges from 0.05 to 1.0 mm for the maxillary sinuses, 0.07 to 0.7 mm for the frontal sinuses, 0.08 to 0.8 mm for the ethmoid sinuses, and 0.07 to 0.7 for the sphenoid sinuses. The thickness of the simple columnar epithelium, including the cilia, is between 30 and 45 µm.

The difference between mucosal and epithelial thickness gives the thickness of the lamina propria a quantity of importance for dosimetry. In the simple columnar epithelium, the thicknesses for the lamina propria implied by the preceding information range from about 10 µm upward to nearly 1 mm. Direct observations of the lamina propria indicate that the thickness lies between 14 and 541 µm.²¹

Mucosal dimensions for the mastoid air cells have been less well studied. Littman et al.³¹ report a single value of 17 µm for the lamina propria in a person who had contracted mastoid carcinoma. In a more complete series of measurements on normal persons and persons exposed to low ^226,228Ra doses, Harris and Schlenker²¹ reported total mucosal thicknesses between 22 and 134 µm, with epithelial thicknesses in the range of 3 to 14 µm and lamina propria thicknesses in the range of 19 to 120 µm.

The normally functioning sinus is ventilated; that is, its ostium or ostia are open, permitting the free exchange of gases between the sinus and nasal cavities. When the sinus becomes unventilated due to ostial closure, the gas composition of the sinus cavity changes and slight overpressure or underpressure may occur.¹³ When radioactive gases (radon) are present, as with persons exposed to ^226,228Ra, there is the potential for a much higher concentration of those gases in the air of the sinus when unventilated than when ventilated. Ventilation of the mastoid air cells occurs through the eustachian tube which normally allows little air to move. Thus, there is a potential for the accumulation of large quantities of radon. Cancer of the paranasal sinuses and mastoid air cells has been associated with ^226,228Ra exposure since the late 1930s⁴³ following the death of a radium-dial painter who had contracted epidermoid carcinoma of the epithelium lining of the ethmoid air cells.³

The natural tumor rate in these regions of the skull is very low, and this aids the identification of etiological agents. Malignancies of the auditory tube, middle ear, and mastoid air cells (ICD 160.1) make up only 0.0085% of all malignancies reported by the National Cancer Institute's SEER program.⁵² Those of the ethmoid (ICD 160.3), frontal (ICD 160.4), and sphenoid (ICD 160.5) sinuses together make up 0.02% of all malignancies, or if the nonspecific classifications, other (ICD 160.8) and accessory sinus, unspecified (ICD 160.9), are added as though all tumors in these groups had occurred in the ethmoid, frontal, or sphenoid sinuses, the incidence would be increased only to 0.03% of all malignancies. In 1977 it was estimated that only 15 people died in the United States from cancers of the auditory tube, middle ear, and mastoid air cells.⁵³ Comparable statistics are lacking for cancers of the ethmoid, frontal, and sphenoid sinuses; but mortality, if scaled from the incidence data, would not be much greater than that caused by cancers of the auditory tube, middle ear, and mastoid air cells.

Carcinomas of the paranasal sinuses and mastoid air cells may invade the cranial nerves, causing problems with vision or hearing^3,23 prior to diagnosis. Littman et al.³¹ have presented a list of symptoms in tabular form gleaned from a study of the medical records of 32 subjects who developed carcinoma of the paranasal sinuses or mastoid air cells following exposure to ^226,228Ra. The most frequent clinical symptoms for paranasal sinus tumors were problems with vision, pain (not specified by location), nasal discharge, cranial nerve palsy, and hearing loss. The most frequent symptoms for mastoid air cell tumors were ear blockage or discharge and hearing loss.

Some 35 carcinomas of the paranasal sinuses and mastoid air cells have occurred among the 4,775 ^226,228Ra-exposed patients for whom there has been at least one determination of vital status. For 31 of the tumors, estimates of skeletal dose can and have been made. Data on tumor locations and histologic type are presented in Table 4-4. No maxillary sinus carcinomas have occurred, but 69% of the tumors have occurred in the mastoids. For tumors of known histologic type, 56% are epidermoid, 34% are mucoepidermoid, and 10% are adenocarcinomas. For the sinuses alone, the distribution of types is 40% epidermoid, 40% mucoepidermoid, and 20% adenocarcinoma, compared with 37, 0, and 24%, respectively, of naturally occurring carcinomas in the ethmoid, frontal, and sphenoid sinuses.⁴ Among all microscopically confirmed carcinomas with known specific cell type in the nasal cavities, sinuses and ear listed in the National Cancer Institute SEER report,⁵² 75% were epidermoid, 1.6% were mucoepidermoid, and 7% were adenocarcinoma. The rarity of naturally occurring mucoepidermoid carcinoma, contrasted with its frequency among ^226,228Ra-exposed subjects, suggests that alpha-particle radiation is capable of significantly altering the distribution of histologic types. The complete absence of other, less-frequent types of naturally occurring carcinoma that represent 16% of the carcinomas of specific cell type in the SEER⁵² study and 39% of the carcinomas in the review by Batsakis and Sciubba⁴ provides further evidence for perturbation of the distribution of carcinoma types by alpha radiation.

TABLE 4-4

Carcinomas of the Paranasal Sinuses and Mastoid Air Cells among Persons Exposed to ^226,228Ra and Currently Under Study at Argonne National Laboratory.

The cause of paranasal sinus and mastoid air cell carcinomas has been the subject of comment since the first published report,⁴³ when it was postulated that they arise ''. . . in the mucosa . . . as result of the local effects of the radon . . . in the expiratory air . . . ." In 1952, Aub et al.³ stated that the origin of these neoplasms in mucosal cells that were well beyond the range of the alpha particles emitted by radium, mesothorium, and their bone-fixed disintegration products is also interesting. Only the beta and gamma rays, which were of low intensity compared to the alpha rays, emitted by these radioactive materials in the adjacent bone could have reached these cells. However, the mucosa may have been irradiated by the alpha rays from the radiothorium that was fixed in the adjacent periosteum. Also, they were continuously subjected to alpha radiation from another source: the radon in expired breath. Restated in more modern terms, the residual range from bone volume seekers (²²⁶Ra and ²²⁸Ra) is too small for alpha particles to reach the mucosal epithelium, but the range may be great enough for bone surface seekers (²²⁸Th), whose alpha particles suffer no significant energy loss in bone mineral;⁷⁸ long-range beta particles and most gamma rays emitted from adjacent bone can reach the mucosal cells, and free radon may play a role in the tumor-induction process. Hasterlik²² and Hasterlik et al.²³ further elucidated the role of radon by postulating that it can diffuse from bone into the essentially closed airspaces of the mastoid air cells and paranasal sinuses and decay there with its daughters, adding an additional dose to the epithelial cells.

A significant role for free radon and the possibly insignificant role for bone volume seekers is not universally acknowledged; the ICRP lumps the sinus and mastoid mucosal tissues together with the endosteal bone tissues and considers that the dose to the first 10 µm of tissue from radionuclides deposited in or on bone is the carcinogenically significant dose, thus ignoring trapped radon altogether and taking no account of the epithelial cell locations which are known to be farther from bone than 10 µm.

The first attempts at quantitative dosimetry were those of Kolenkow³⁰ who presented a detailed discussion of frontal sinus dosimetry for two subjects, one with and one without frontal sinus carcinoma. In the subject with carcinoma, he observed a hot layer of bone beginning about 2 µm from the surface and extending inward a distance greater than the alpha-particle range. The radium concentration in this layer was 50 to 75 times the mean concentration for the whole skeleton. In the subject without carcinoma, the measured radium concentration in the layer adjacent to the bone surface was only about 3 times the skeletal average. Kolenkow³⁰ presented his results as depth-dose curves for the radiation delivered from bone but made no comment on epithelial cell location. The depth dose for radon and its daughters in the frontal sinus of the subject with carcinoma was based on a direct measurement of radon activity in the unaffected frontal sinus at the time surgery was performed on the diseased sinus. The calculated dose from this source was much less than the dose from bone. Based on Kolenkow's work,³⁰ Evans et al.¹⁶ reported a cumulative dose of 82,000 rad to the mucous membrane at a depth of 10 µm for the subject with carcinoma. Kolenkow's work³⁰ illustrated many of the complexities of sinus dosimetry and emphasized the rapid decrease of dose with depth in the mucous membrane.

The first explicit description of the structure of the sinus and mastoid mucosa in the radium literature is probably that of Hasterlik,²² who described it as "thin wisps of connective tissue," overlying which "is a single layer of epithelial cells. . . ." He placed the total thickness of connective tissue plus epithelium at between 5 and 20 µm. A clear implication of these data is that the connective tissue in the mastoid is thinner than the connective tissue in the paranasal sinuses. In a dosimetric study, Schlenker⁷³ confirmed this by determining the frequency with which the epithelium lay nearer to or farther from the bone surface than 75 µm, at which level more than 75% of the epithelial layer in the mastoids would be irradiated. Commenting on the mucosal thickness data of Ash and Raum,² Littman et al.³¹ observed: "If the dimensions of the sinus walls are applicable to the radium cases, it would appear that only a relatively sparse population of epithelial cells in the submucosal glands of the paranasal sinuses would receive significant dose from alpha particles originating in bone."

Equations for the dose rate averaged over depth, based on a simplified model of alpha-particle energy loss in tissue, were presented by Littman et al.³¹ for dose delivered by radium in bone and by radon and its daughters in an airspace with a rectangular cross section. They also presented an equation for depth dose from radon and its daughters in the airspace for the case of a well-ventilated sinus, in which the radon concentration was equal to the radon concentration in exhaled breath. For the 27 subjects for whom radium body burden information was available, they estimated that, for airspace thicknesses of 0.5 to 2 cm, the dose from radon and its daughters averaged over a 50-µm-thick mucous membrane would be 2 to 5% of the average dose from ²²⁶Ra in bone. Clearly, under these assumptions, dose from radon and its daughters in the airspaces would be of little radiological significance.

The quantitative impact of cell location on dosimetry was emphasized by Schlenker⁷⁵ who focused attention on the relative importance of dose from radon and its daughters in the airspaces compared to dose from radium and its daughters in bone. He emphasized that current recommendations of the ICRP make no clear distinction between the locations of epithelial and endosteal cells and leave the impression that both cell types lie within 10 µm of the bone surface; this leads to large overestimates of the dose to epithelial cells from bone.

In a more complete development, Schlenker⁷³ investigated the dosimetry of sinus and mastoid epithelia when ²²⁶Ra or ²²⁸Ra was present in the body. He used the same assumptions about linear energy transfer as Littman et al.;³¹ adopted a spherical shape for the air cavities; and considered air cavity diameters from 0.2 mm, representing small mastoid air cells, up to 5 cm, representing large sinuses. He took into account the dose rate from ²²⁶Ra or ²²⁸Ra in bone, the dose rate from ²²²Rn or ²²⁰Rn in the airspaces, the impact of ventilation and blood flow on the residence times of these gases in the airspaces, measured values for the radioactivity concentrations in the bones of certain radium-exposed patients, and determined expected values for radon gas concentrations in the airspaces. For five subjects on whom he had autoradiographic data for the ²²⁶Ra specific activity in bone adjacent to the mastoid air cells, the dose rate at death from ²²²Rn and its daughters in the airspaces exceeded the dose rate from ²²⁶Ra and its daughters in bone. On average, the dose rate from airspaces was about 4 times that from bone.

He also estimated dose rates for situations where there were no available autoradiographic data. The dose rate from the airspaces exceeded the dose rate from bone when ²²⁶Ra or ²²⁸Ra was present in the body except in one situation. For ²²⁸Ra the dose rate from the airspace to the mastoid epithelium was about 45% of the dose rate from bone. These results are in marked contrast to those of Kolenkow³⁰ and Littman et al.³¹ Under Schlenker's⁷³ assumptions, the airspace is the predominant source of dose, with the exception noted, whether or not the airspace is ventilated.

Working from various radium-exposed patient data bases, several authors have observed that carcinomas of the paranasal sinuses and mastoid air cells begin to occur later than bone tumors.^16,18,66,71 In the latest tabulation of tumor cases,¹ the first bone tumor appeared 5 yr after first exposure, and the first carcinoma of the paranasal sinuses or mastoid air cells appeared 19 yr after first exposure; among persons for whom there was an estimate of skeletal radiation dose, the first tumors appeared at 7 and 19 yr, respectively. The frequency distribution for appearance times shows a heavy concentration of paranasal sinus and mastoid carcinomas with appearance times of greater than 30 yr. For bone tumors there were approximately equal numbers with appearance times of less than or greater than 30 yr.⁶⁷ Based on the most recent summary of data, 32 bone tumors occurred with appearance times of less than 30 yr among persons with known radiation dose and 29 tumors had occurred with appearance times of 30 yr or greater. Within the same group, four carcinomas occurred with appearance times equal to or greater than 30 yr. Unless there is a bias in the reporting of carcinomas, it is clear that carcinomas are relatively late-appearing tumors.

Rowland et al.⁶⁶ plotted and tabulated the appearance times of carcinomas for five different dosage groups. On the basis of minimum and median appearance times, they concluded that the appearance times do not change with dose. Rowland et al.⁶⁷ performed a dose-response analysis of the carcinoma data in which the rate of tumor occurrence (carcinomas per person-year at risk) was determined as a function of radium intake. The linear relationship that provided the best fit to the data predicted a tumor rate lower than the rate that had been observed recently, and led the authors to suggest that the incidence at long times after first exposure may be greater than the average rate observed thus far.

An analysis of the tumor appearance time data for carcinomas based on hazard plotting has been as employed by Groer and Marshall²⁰ to analyze bone tumor rate in persons exposed to high doses from radium. The data are subdivided into three groups based on the ²²⁶Ra intake. ²²⁸Ra intake was excluded because it was assumed that ²²⁸Ra is ineffective for the production of these carcinomas. Data points fall along a straight line when the tumor rate is constant. The intersection of the line with the appearance time axis provides an estimate of the minimum appearance time. The analysis shows that the minimum appearance time varies irregularly with intake (or dose) and that the rate of tumor occurrence increases sharply at about 38 yr after first exposure for intakes of greater than 470 µCi and may increase at about 48 yr after first exposure for intakes of less than 260 µCi.

As of the 1980 follow-up, no carcinomas of the paranasal sinuses and mastoid air cells had occurred in persons injected with ²²⁴Ra, although Mays and Spiess⁴⁶ estimated that five carcinomas would have occurred if the distribution of tumor appearance times were the same for ²²⁴Ra as for ^226,228Ra.

Carcinomas of the frontal sinus and the tympanic bulla, a portion of the skull comparable to the mastoid region in humans, have appeared in beagles injected with radium isotopes and actinides. Based on epizootiological studies of tumor incidence among pet dogs, Schlenker⁷³ estimated that 0.06 tumors were expected for 789 beagles from the University of Utah beagle colony injected with a variety of alpha emitters, while five tumors were observed. Three of the five tumors were induced by actinides that have no gaseous daughter products. Their induction, therefore, cannot be influenced by dose from the airspace as can the induction of carcinomas by ²²⁶Ra in humans. The beagle data demonstrate that a gaseous daughter product is not essential for the induction of sinus and mastoid carcinomas, while Schlenker's⁷³ dosimetric analysis and the epidemiological data^16,67 indicate that it is an important factor in human carcinoma induction. The conclusion from this and information on tissue dimensions is that the sinuses, and especially the mastoids, are at risk from alpha emitters besides ²²⁶Ra, but that the risk may be significantly lower than that from ²²⁶Ra and its decay products.

Rowland et al.⁶⁷ have reported the only separate analyses of paranasal sinus and mastoid carcinoma incidence. As the response variable, they used carcinomas per person-year at risk and regressed it against a measure of systemic intake of ²²⁶Ra and against average skeletal dose. They fit mathematical functions of the general form:

in which all three coefficients (α, β , γ) were allowed to vary or one or more of the coefficients were set equal to zero. In this expression, C is the natural carcinoma rate and D is the systemic intake or mean skeletal dose. The best fit of response against systemic intake was obtained for the functional form I = C + αD, obtained from Equation 4-21 by setting β = γ = 0. The poorest fit, and one that is unacceptable according to a chi-squared criterion, was obtained for I = C + β D². All other functional forms gave acceptable fits. When persons that had entered the study after exhumation were excluded from the analysis, in an effort to control selection bias, all six forms of the general function gave acceptable fits to the data. The exclusion of exhumed subjects removed from analysis 23 of the 759 individuals in the population and 1 of the 21 carcinomas that had occurred among them. This change had no effect on the fitted value of α, the free parameter in the linear dose-response function.

The analysis of response as a function of ²²⁶Ra dose was conducted with exhumed cases included. The best fit was obtained for the functional form I =(C + αD) exp(-γD), an unacceptable fit was obtained for I = C + β D², and all other forms provided acceptable fits.

The linear functions obtained by Rowland et al.⁶⁷ were:

where D _i is ²²⁶Ra intake, and D _s is ²²⁶Ra skeletal dose. In the data analyses that lead to these equations, a 10-yr latent period is assumed for carcinoma induction. This latent period must be included when the equations are applied to risk estimation. For example, if a person is exposed to ²²⁶Ra at time zero, the person is not considered to be at risk for 10 yr; the total number of carcinomas expected to occur among N people with identical systemic intakes D _i is IN (t - 10) for t > 10 yr and 0 for t < 10 yr. This is also true for N people, all of whom accumulate a skeletal dose D _s.

The analysis of Rowland et al.⁶⁷ assumes that tumor rate is constant with time for a given intake D _i, and when based on skeletal dose assumes that tumor rate is constant for a given dose D _s. The analysis also yields good fits to the data. It should be noted that if tumor rate were constant for a given dose, it could not be constant for a given intake because the dose produced by a given intake is itself a function of time; therefore, the tumor rate would be time dependent. The success achieved in fitting dose-response functions to the data, both as a function of intake and of dose, indicates that the outcome is not sensitive to assumptions about tumor rate. No firm conclusions about the constancy or nonconstancy of tumor rate should be drawn from this dose-response analysis. Recall that the preceding discussion of tumor appearance time and rate of tumor appearance indicated that tumor rate increases with time for some intake bands, verifying a suggestion by Rowland et al.⁶⁷ made in their analysis of the carcinoma data.

The subjects used in this analysis were all women employed in the radium-dial-painting industry at an average age of about 19 yr. There is no assurance that women exposed at a greater age or that men would have yielded the same results. This represents a nonquantifiable uncertainty in the application of the preceding equations to risk estimation.

The statistical uncertainty in the coefficient α is determined principally by the variance in the high-dose data, that is, at exposure levels for which the observed number of tumors is nonzero. In this analysis, there were one or more tumors in the six intake groups with intakes above 25 µCi and no tumors observed in groups with intakes below 25 µCi. Therefore, calculations of the uncertainty of risk estimates from the standard deviation will be accurate above 25 µCi but may be quite inaccurate and too small below 25 µCi. Schlenker⁷⁴ examined the uncertainties in risk estimates for bone tumor induction at low intakes and found it to be much greater than would be determined from the standard deviations in fitted risk coefficients. The analysis was not carried out for carcinoma risk, but the conclusions would be the same.

Leukemia

Leukemia has not often been seen in the studies of persons who have acquired internally deposited radium. Nevertheless, the discussion of leukemia as a possible consequence of radium exposure has appeared in a number of published reports.

Martland,⁴² summarizing his studies of radium-dial painters, mentioned the development of anemias. He also described the development of leukopenia and anemia, which appeared resistant to treatment.

Evans¹⁵ listed possible consequences of radium acquisition, which included leukemia and anemia. However, no mention of such cases appear in his report.

In a report by Finkel et al.,¹⁸ mention is made of seven cases of leukemia and aplastic anemia in a series of 293 persons, most of whom had acquired radium between 1918 and 1933. Five of these cases of leukemia were found in a group of approximately 250 workers from radium-dial painting plants in Illinois. The radium content in the bodies of 185 of these workers was measured. There were three cases of chronic myeloid leukemia (CML) and one of chronic lymphocytic leukemia (CLL). Because CLL is not considered to be induced by radiation, the latter case was assumed to be unrelated to the radium exposure. The remaining two cases were aplastic anemias; these latter two cases and one of the CML cases were not available for study, and hence no measurements of radium content in the workers' bodies were available.

In an additional group of 37 patients who were treated with radium by their personal physicians, two blood dyscrasias were found. One of these was panmyelosis, and the other was aplastic anemia; the radium measurements for these two cases showed body contents of 10.5 and 10.7 µCi, respectively. The average skeletal doses were later calculated to be 23,000 and 9,600 rad, respectively, which are rather substantial values.

Finkel et al.¹⁸ concluded that the appearance of one case of CML in 250 dial workers, with about 40 yr of follow-up time, would have been above that which was expected. Two cases, by implication, might be considered significant.

In a review of the papers published in the United States on radium toxicity, and including three cases of radium exposure in Great Britain, Loutit³⁴ made a strong case "that malignant transformation in the lymphomyeloid complex should be added to the accepted malignancies of bone and cranial epithelium as limiting hazards from retention of radium." He pointed out that the reports of Martland^41–43 describe a regenerative leucopenic anemia, and he stated that "this syndrome has features of atypical (aleukemic) leukemia or myelosclerosis or both."

It should be noted, however, that the early cases of Martland were all characterized by very high radium burdens. The British patients that Loutit described³⁴ also may have experienced high radiation exposures; two were radiation chemists whose radium levels were reported to fall in the range of 0.3 to 0.5 µCi, both of whom probably had many years of occupational exposure to external radiation. The third patient was reported to contain 4–5 µg of radium.

Following the consolidation of the U.S. radium cases into a single study at the Argonne National Laboratory, Polednak⁵⁷ reviewed the mortality of women first employed before 1930 in the U.S. radium-dial-painting industry. This study examined a cohort of 634 women who had been identified by means of employment lists or equivalent documents. This cohort was derived from a total of about 1,400 pre-1930 radium-dial workers who had been identified as being part of the radium-dial industry of whom 1,260 had been located and were being followed up at Argonne. By 1954, when large-scale studies of the U.S. radium cases were initiated, 521 of the cohort of 634 women were still alive, and 360 of them had whole-body radium measurements made after that date while they were still living.

In the cohort of 634 women, death certificates indicated that there were three cases attributed to leukemia and aleukemia and four more to blood and blood-forming organs; both were above expectations. When the study was restricted to the 360 measured cases, one case of leukemia was found in a woman with a radium intake greater than 50 µCi. Similarly, only one death attributable to diseases of the blood, acquired hemolytic anemia, was found for a person with a very low radium intake.

Stebbings et al.⁸⁹ published results of a mortality study of the U.S. female radium-dial workers using a much larger data base. This study included 1,285 women who were employed before 1930. A comparison study included 1,185 women employed between 1930 and 1949, when radium contamination was considerably lower.

In this enlarged study, three cases of leukemia were recorded in the pre-1930 population, which yielded a standard mortality ratio of 73. All of these cases occurred among 293 women employed in Illinois; none were recorded among the employees from radium-dial plants in other states. An additional three cases were found in the 1930–1949 cohort, yielding a standard mortality ratio of 221. These authors concluded that there was no relationship between radium level and the occurrence of leukemia.

The most inclusive and definitive study of leukemia in the U.S. radium-dial workers was published by Spiers et al.⁸³ By including all the dial workers, male and female, who entered the industry before 1970, a total of 2,940 persons who could be located, they were able to document a total of 10 cases of leukemia. A total of 9.2 cases would be expected to occur naturally in such a population. Table 4-5, based on their report, illustrates their results.

TABLE 4-5

Incident Leukemia in Located Radium Workers.

Included in the above summary are four cases of chronic lymphocytic or chronic lymphatic leukemia. Spiers et al.⁸³ note that this number from a total of 10 is not dissimilar from the 3.6 expected in the general population. They conclude that the incidence of myeloid and other types of leukemia in this population is not different from the value expected naturally.

When the U.K. radium-luminizer study for the induction of myeloid leukemia is examined,⁵ it is seen that among 1,110 women there are no cases to be found. The expected number, however, is only 1.31.

The above results, based on observations of several thousand individuals over periods now ranging well over 50 yr, make the recent report by Lyman et al.³⁵ on an association between radium in the groundwater of Florida and the occurrence of leukemia very difficult to evaluate. Florida has substantial deposits of phosphate, and this ore contains ²³⁸U, which in turn produces ²²⁶Ra and ²²²Rn. The radium from this ore evidently finds its way into the groundwater supplies. By measuring the radium content of 50 private wells in 27 selected counties, the counties were divided into 10 low-exposure and 17 high-exposure groups. The high-exposure group was further divided into three graded groups. These divisions were made on the basis of the number of these private wells in each county that contained more than 5 pCi/liter of water.

Lyman et al.³⁵ show a significant association between leukemia incidence and the extent of groundwater contamination with radium. The majority of the leukemias were acute myeloid leukemias. Further, a dose-response relationship is suggested for total leukemia with increasing levels of radium contamination.

Lyman et al.³⁵ do not claim, however, to have shown a causal relationship between leukemia incidence and radium contamination. They point out that there is no information on individual exposure to radium from drinking water, nor to other confounding factors. Since leukemia rates are not elevated in the radium-dial worker studies, where the radium exposures ranged from near zero to many orders of magnitude greater than could be attributed to drinking water, it is difficult to understand how radium accounts for the observations in this Florida study.

In summary, the evidence indicates that acquisition of very high levels of radium, leading to long-term body contents of the order of 5 µCi or more, equivalent to systemic intakes of the order of several hundred microcuries, resulted in severe anemias and aleukemias. However, at lower radium intakes, such as those experienced by the British luminizers and the bulk of the U.S. radium-dial workers, incorporated ²²⁶Ra does not appear to give rise to leukemia.

The 3.62-day half-life of ²²⁴Ra results in a prompt, short-lived pulse of alpha radiation; in the case of the German citizens injected with this radium isotope, this pulse of radiation was extended by repeated injections. Nevertheless, the time that bone and adjacent tissues were irradiated was quite short in comparison to the irradiation following incorporation of ²²⁶Ra and ²²⁸Ra by radium-dial workers. In spite of these differences, ²²⁴Ra has been found to be an efficient inducer of bone cancer.

In the case of leukemia, the issue is not as clear. Leukemia has been seen in the Germans exposed to ²²⁴Ra, but only at incidence rates close to those expected in unexposed populations. When an excess has occurred, there exist confounding variables.

Mays et al.⁵⁰ reported on the follow-up of 899 children and adults who received weekly or twice-weekly intravenous injections of ²²⁴Ra, mainly for the treatment of tuberculosis and ankylosing spondylitis. While five cases of leukemia were observed among 681 adults who received an average skeletal dose of 206 rad, none were observed among 218 1 – to 20-yr-olds at an average skeletal dose of 1,062 rad. The expected number of leukemias for the adult group was two, but the authors point out that the drugs often taken to suppress the pain associated with ankylosing spondylitis are suspected of inducing the acute forms of leukemia. Four of the five leukemias occurred in patients with ankylosing spondylitis; two were known to be acute; it is not known whether the other three were acute or chronic.

It is evident that leukemia was not induced among those receiving ²²⁴Ra before adulthood, in spite of the high skeletal doses received and the postulated higher sensitivity at younger ages. There may be an excess of leukemia among the adults, but the evidence is weak.

Wick et al.⁹⁵ reported on another study of Germans exposed to ²²⁴Ra. While the report of Mays et al.⁵⁰ dealt with persons injected with ²²⁴Ra between 1946 and 1950, the study of Wick et al.⁹⁵ examined the consequences of lower doses as a treatment for ankylosing spondylitis and extended from 1948 to 1975. The average skeletal dose to a 70-kg male was stated to be 56 rad. There were 1,501 exposed cases and 1,556 ankylosing spondylitis controls. Each group consisted of about 90% males. There were 11 bone marrow failures in the exposed group, and only 4 in the control group. Similarly, there were six leukemias in the exposed group versus five in the control group. All five leukemias in the control group were acute forms, while three in the exposed group were chronic myeloid leukemia. The authors drew no conclusions as to whether the leukemias observed were due to ²²⁴Ra, to other drugs used to treat the disease, or were unrelated to either.

This population has now been followed for 34 yr; the average follow-up for the exposed group is about 16 yr. A total of 433 members of the exposed group have died, leaving more than 1,000 still alive. It may be some time before this group yields a clear answer to the question of radium-induced leukemia. At this time, it is clear that it is not a primary consequence of radium deposited in human bones.

Thus, while leukemia and diseases of the blood-forming organs have been seen following treatment with ²²⁴Ra, it is not clear that these are consequences of the radiation insult or of other treatments experienced by these patients. The extremely high radiation doses experienced by a few of the radium-dial workers were not repeated with ²²⁴Ra, so clear-cut examples of anemias following massive doses to bone marrow are lacking.

Radium in Water

Since uranium is distributed widely throughout the earth's crust, its daughter products are also ubiquitous. As a consequence, many sources of water contain small quantities of radium or radon. In the United States there have been at least three attempts to determine whether the populations that drink water containing elevated levels of radium had different cancer experience than populations consuming water with lower radium levels.

A cooperative research project conducted by the U.S. Public Health Service and the Argonne National Laboratory made a retrospective study of residents of 111 communities in Iowa and Illinois who were supplied water containing at least 3 pCi/liter by their public water supplies. Control cities where the radium content of the public water supply contained less than 1 pCi/liter were matched for size with the study cities. A total of almost 908,000 residents constituted the exposed population; the mean level of radium in their water was 4.7 pCi/liter.

The final report of this study by Petersen et al.⁵⁶ reported on the number of ''deaths due in any way to malignant neoplasm involving bone." They found that, for the period 1950–1962, the age- and sex-adjusted rate for the radium-exposed group was 1.41/100,000/yr. The rate for the control group was 1.14; the probability of such a difference occurring by chance alone was reported as 8 in 100.

However, Petersen⁵⁵ wrote an interim report for a review board constituted to advise on a proposal for continued funding for this project. This report indicates that the age- and sex-adjusted osteosarcoma mortality rate for the total white population in the communities receiving elevated levels of radium for the period 1950–1962 was 6.2/million/yr; that of the control population was 5.5. The probability of such a difference occurring by chance was 51%. The authors concluded that "no significant difference could be detected between the osteosarcoma mortality rate in towns with water supplies having elevated levels of ²²⁶Ra and matched control towns." The complexity of the problem is illustrated by their findings for Chicago. Although this city draws its water from Lake Michigan, where the radium concentration is reported as 0.03 pCi/liter, the age- and sex-adjusted osteosarcoma mortality rate was 6.3/million/yr, which is larger than that found for the towns with elevated radium levels in their water.

The mobility of populations in this country, the inability to document actual radium intakes, and the fact that water-softening devices remove radium from water all tend to make studies of this nature very difficult to evaluate.

A pair of studies relating cancer to source of drinking water in Iowa were reported by Bean and coworkers.^6,7 The first of these examined the source of water, the depth of the well, and the size of the community. This study was aimed at the role, if any, of trihalomethanes resulting from the disinfection of water by chlorination. The second, which used the deep-well data from the prior study, examined cancer incidence as a function of radium content of the water. Twenty-eight towns met the three criteria for the second study: a population between 1,000 and 10,000, water is obtained solely from wells greater than 500 ft (152 m) deep, and no water softening. These 28 towns had a total population of 63,689 people in 1970.

When the water supplies were divided into three groups levels of 0–2, 2–5, and > 5 pCi of ²²⁶Ra per liter and the average annual age-adjusted incidence rates were examined for the period 1969–1978 (except for 1972), certain cancers were found to increase with increasing radium content. These were bladder and lung cancer for males and breast and lung cancer for females. Their data, plus the incidence rates for these cancers for all Iowa towns with populations 1,000 to 10,000 are shown in Table 4-6. When examined in this fashion, questions arise. There is no doubt that male and female lung cancers appear to increase with an increase in the radium content of the water, but in the case of female lung cancers the levels were never as great as observed for those who drank surface water. A similar situation exists for female breast cancer. For male bladder cancer only, the highest radium level produced a higher cancer rate than was observed for those consuming surface water. Were it not for the fact that these cancers were not seen at radium intakes hundreds to thousands of times greater in the radium-dial painter studies, they might throw suspicion on radium. However, it is difficult to accept this hypothesis without an explanation of the lesser number of cancers found at higher radium intakes.

TABLE 4-6

Cancer Incidence Rate among Persons Exposed to Different Concentrations of Radium in Drinking Water.

In summary, there are three studies of radium in drinking water, one of which found elevated "deaths due in any way to malignant neoplasm involving bone," the second found elevated incidences of bladder and lung cancer in males and lung and breast cancer in females, and the third found elevated rates of leukemia. None of these findings are in agreement with the long-term studies of higher levels of radium in the radium-dial workers.

Risk Estimation

There is little evidence for an age or sex dependence of the cancer risk from radium isotopes, provided that the age dependence of dose that accompanies changes in body and tissue masses is taken into account. With the present state of knowledge, a single dose-response relationship for the whole population according to isotope provides as much accuracy as possible. For ²²⁴Ra, ²²⁶Ra, and ²²⁸Ra the best-available relationships are based on different measures of exposure: absorbed skeletal dose for ²²⁴Ra and systemic intake for ²²⁶Ra and ²²⁸Ra. Simple prescriptions for the skeletal dose from ²²⁴Ra as a function of injection level have been given by Spiess and Mays⁸⁵ and can be used to estimate skeletal dose from estimated systemic intake. Because all of the data analysis for ²²⁴Ra has been based on prescription of dose given by Spiess and Mays,⁸⁵ it is important that it be followed in applications of ²²⁴Ra dose-response relationships for the estimation of cancer risk in the general population or in case of occupational or therapeutic exposure. Shifting to a different algorithm for dose calculation would, at a minimum, require demonstration that the new algorithm gives the same numerical values for dose as the Spiess and Mays⁸⁵ algorithm for subjects of the same age and sex. The alternative is to reanalyze all of the data on tumor induction for ²²⁴Ra by using the new algorithm before it is applied it to dose calculations for risk estimation in a population group different from the subjects in the study by Spiess and Mays.⁸⁵

For ingested or inhaled ²²⁴Ra, a method for relating the amount taken in through the diet or with air to the equivalent amount injected in solution is required. The ICRP models for the gastrointestinal tract and for the lung provide the basis for establishing this relationship. A necessary first step for the estimation of risk from any route of intake other than injection is therefore to apply these models.

A similar issue exists for ²²⁶Ra and ²²⁸Ra. Here the available dose-response relationships are presented in terms of the number of microcuries that reach the blood. Intake by inhalation or ingestion must again account for transfer of radium across the intestinal or pulmonary membranes when the ICRP models are used.

For ²²⁴Ra the dose-response relationship gives the lifetime risk of bone cancer following an exposure of up to a few years' duration. Because bone cancer is an early-appearing tumor, the risk, so far as is now known, disappears within 25 yr after exposure. It peaks about 5 yr after exposure following the passage of a minimum latent period. Thus, the absence of information on the tumor probability as a function of person-years at risk is not a major limitation on risk estimation, although a long-term objective for all internal-emitter analyses should be to reanalyze the data in terms of a consistent set of response variables and with the same dosimetry algorithm for both ²²⁴Ra and for ²²⁶Ra and ²²⁸Ra. When the time dependence of bone tumor appearance following ²²⁴Ra exposure is considered an essential component of the analysis, then an approximate modification of the dose-response relationship can be made by taking the product of the dose-response equation and an exponential function of time to represent the rate of tumor appearance:

where F(D) is the lifetime risk, as specified by the analyses of Spiess and Mays⁸⁵ and r is a coefficient based on the time of tumor appearance for juveniles and adults in the ²²⁴Ra data analyses. The half-life for tumor appearance is roughly 4 yr in this data set, giving an approximate value for r of 0.18/yr. For t less than 5 yr, M(D,t) is essentially 0 because of the minimum latent period. Thereafter, tumors appear at the rate M(D,t).

The age structure of the population at risk and competing causes of death should be taken into account in risk estimation. An ideal circumstance would be to know the dose-response relationships in the absence of competing causes of death and to combine this with information on age structure and age-specific mortality for the population at large. With the analyses presently available, only part of this prescription can be achieved. An approximate approach would be to take the population as a function of age and exposure and apply the dose-response relationship to each age group, taking into account the projected survival for that age group in the coming years. At the low exposures that occur environmentally and occupationally, exposure to radium isotopes causes only a small contribution to overall mortality and would not be expected to perturb mortality sufficiently to distort the normal mortality statistics. Also, mortality statistics as they now exist include the effect of environmental exposures to radium isotopes.

Table 4-7 illustrates the effect, assuming that one million U.S. white males receive an excess skeletal dose of 1 rad from ²²⁴Ra at age 40. The excess death rate due to bone cancer for t > 5 yr is computed from:

TABLE 4-7

Effect of Single Skeletal Dose of 1 rad from ²²⁴Ra Received by 1,000,000 U.S. White Males at Age 40.

This assumes the ²²⁴Ra dose-response analyses described above and further assumes that tumors are fatal in the year of occurrence. After 25 yr, there would be 780,565 survivors in the absence of excess exposure to ²²⁴Ra and 780,396 survivors with 1 rad of excess exposure at the start of the follow-up period, a difference of 169 excess deaths/person-rad, which is about 15% less than the lifetime expectation of 200 × 10^-6/person-rad calculated without regard to competing risks.

If there were a continuous exposure of 1 rad/yr, the tumor rate would rise to an asymptotic value. If this were substituted for the tumor rate caused by ²²⁴Ra exposure in Table 4-7 and the survival rate of those exposed to ²²⁴Ra were adjusted to the corresponding value (0.9998), survival in the presence of ²²⁴Ra exposure after 25 yr would be 777,293, with 3,272 deaths attributable to the ²²⁴Ra exposure.

Calculations for ²²⁶Ra and ²²⁸Ra are similar to the calculation with the asymptotic tumor rate for ²²⁴Ra. For ²²⁶Ra and ²²⁸Ra the constant tumor rates given by Rowland et al.⁶⁸ as functions of systemic intake are computed for the intake of interest, and the results are worked out with a table such as Table 4-7. For continuous intake with the dose-squared exponential function for bone sarcoma induction, it is necessary to decide whether to add the cumulative dose and then take the square or to take the square for each annual increment of dose. Taking the former choice, it is implied that the doses given at different times interact; with the latter choice it is implied that the doses act independently of one another. On the microscale the chance of a single cell being hit more than once diminishes with dose; this would argue for the independent action of separate dose increments and the squaring of separate dose increments before the addition of risks. In the model of bone tumor induction proposed by Marshall and Groer,³⁸ however, two hits are required to cause transformation. This argues for the interaction of doses and in the extreme case for squaring the cumulative dose. Unless bone cancer induced by ²²⁶Ra and ²²⁸Ra is a pure, single-hit phenomenon, some interaction of dose increments is expected, although perhaps it is a less strong interaction than is consistent with squaring the total accumulated intake when intake is continuous.

The advantage of using a tabular form for the calculation of the effect of radiation is that it provides a general procedure that can be applied to more complex problems than the one illustrated above. With environmental radiation, in which large populations are exposed, a spectrum of ages from newborn to elderly is represented. Knowing the death rate as a function of time for each starting age then allows the impact of radiation exposure to be calculated for each age group and to be summed for the whole population. The use of a table for each starting age group provides a good accounting system for the calculation. The same goals can be achieved if normal mortality is represented by a continuous function and radiation-induced mortality is so represented, as for ²²⁴Ra above, and the methods of calculus are used to compute the integrals obtained by the tabular method.

Summary and Recommendations

As documented above, research on radium and its effects has been extensive. With continued research the full fruits of these labors in terms of lifetime risk estimates for ²²⁶Ra and other long-half-life alpha-emitters which are deposited in bone should be realized. In the case of ²²⁴Ra, the relatively short half-life of the material permits an estimation of the dose to bone or one that is proportional to that received by the cells at risk. Correspondingly, relatively simple and complete dose-response functions have been developed that permit numerical estimates of the lifetime risk, that is, about 2 × 10^-2/person-Gy for bone sarcoma following well-protracted exposure. In the case of the longer-half-life radium isotopes, the interpretation of the cancer response in terms of estimated dose is less clear. The dose is delivered continuously over the balance of a person's lifetime, with ample opportunity for the remodeling of bone tissues and the development of biological damage to modulate the dose to critical cells. Deposition (and redeposition) is not uniform and tissue reactions may alter the location of the cells and their number and radiosensitivity. Therefore, estimates of the cumulative average skeletal dose may not be adequate to quantitate the biological insult. Investigation of other dosimetric approaches is warranted.

Equally important is ensuring the availability of information on the rate at which tumors have occurred in the populations at risk. Hazard functions which consider the temporal appearance of tumors have shown some promise for delineating the kinetics of radium-induced bone cancers, and may provide insight into the temporal pattern of the effective dose. Combining this information with results observed with ²²⁴Ra may lead to the development of a general model for bone cancer induction due to alpha-particle emitters.

Further efforts to refine dose estimates as a function of time in both man and animals will facilitate the interpretation of animal data in terms of the risks observed in humans. As indicated in Annex 7A, the radium-dial painter data can be a useful source of information for extrapolating to man the risks from transuranic elements that have been observed in animal studies. A more complete description of the radium-dial painter data and parallel studies with radium in laboratory animals, particularly the rat, would do much to further such efforts.

The committee believes a balanced program of radium research should include the following elements.

• The bone-cancer risk appears to have been completely expressed in the populations from the 1940s exposed to ²²⁴Ra and nearly completely expressed in the populations exposed to ²²⁶Ra and ²²⁸Ra before 1930; the bone-cancer risk data from the two epidemiological studies should be integrated and analyzed with newer statistical methods to extend the usefulness of human data. The committee recommends that these studies continue to include dosimetric evaluation, especially at the tissue and cellular level, and evaluation of uncertainties from all sources.
• The committee recommends that the follow-up studies of the patients exposed to lower doses of ²²⁴Ra since the 1940s now in progress in Germany and of similar groups of patients exposed to ²²⁶Ra and ²²⁸Ra should continue. The detection of bone cancer or sinus and mastoid cancer at dose levels comparable to those encountered in occupational exposures would significantly reduce the uncertainties of bone-cancer risk estimation at low dose levels.
• Research should continue on the cells at risk for bone-cancer induction, on cell behavior over time, including where the cells are located in the radiation field at various stages of their life cycles, on tissue modifications which may reduce the radiation dose to the cells, and on the time behavior and distribution of radioactivity in bone. Meaningful estimates of tissue and cellular dose obtained by these efforts will provide a quantitative linkage between human and animal studies and cell transformation in vitro.
• The sinus and mastoid carcinomas in persons exposed to ²²⁶Ra and ²²⁸Ra are produced largely by the action of ²²²Rn and its progeny; continued study may offer insights into the effects of occupational and environmental radon. The dosimetry of the mastoid air cell system is much simpler than the dosimetry of the bronchial tree; the mastoid mucosa may be the respiratory tissue for which the epithelial structure may permit accurate target cell dose estimates so that the risk to epithelial tissues per unit dose and the specific energy that has an impact on cells can be determined; this may improve our estimation of the carcinogenic risk in the epithelium of the respiratory tract.

References

1.

Argonne National Laboratory, Environmental Research Division. 1984. Annual Report No. ANL-84-103. Argonne, Ill.: Argonne National Laboratory.

2.

Ash, J. E., and M. Raum. Undated. An Atlas of Otolaryngic Pathology, 4th ed. New York: Armed Forces Institute of Pathology.

3.

Aub, J. C., R. D. Evans, L. H. Hempelmann, and H. S. Martland. 1952. The late effects of internally deposited radioactive materials in man. Medicine 31:221–329. [PubMed: 12982323]

4.

Batsakis, J. G., and J. J. Sciubba. 1985. Pathology. Pp. 74–113 , in Surgery of the Paranasal Sinuses, A. Blitzer, editor; , W. Lawson, editor; , and W. H. Friedman, editor. , eds. Philadelpha: W. B. Saunders.

5.

Baverstock, K. F., and D. G. Papworth. 1986. The U.K. radium luminiser survey: Significance of a lack of excess leukemia. Pp. 22–26 in The Radiobiology of Radium and Thorotrast, W. Gössner, editor; , G. B. Gerber, editor; , U. Hagen, editor; , and A. Luz, editor. , eds. Munich, West Germany: Urban and Schwarzenberg.

6.

Bean, J. A., P. Isaacson, W. J. Hausler, and J. Kohler. 1982. Drinking water and cancer incidence in Iowa. I. Trends and incidence by source of drinking water and size of municipality. Am. J. Epidemiol. 116:912–923. [PubMed: 7148817]

7.

Bean, J. A., P. Isaacson, R. M. Hahne, and J. Kohler. 1982. Drinking water and cancer incidence in lowa. II. Radioactivity in drinking water. Am. J. Epidemiol. 116:924–932. [PubMed: 7148818]

8.

Boege, K. 1902. Zur Anatomie der Stirnhohlen, Koniglichen Anatomischen Institut za Konigsberg Nr. 35. Konigsberg in Preussen: Hartangsche Buchdruckerei.

9.

Chemelevsky, D., A. M. Kellerer, H. Spiess, and C. W. Mays. 1986. A proportional hazards analysis of bone sarcoma rates in German radium-224 patients. Pp. 32–37 in The Radiobiology of Radium and Thorotrast, W. Gössner, editor; , G. B. Gerber, editor; , U. Hagen, editor; , and A. Luz, editor. , eds. Munich, West Germany: Urban and Schwarzenberg.

10.

Chiang, C. L. 1968. Introduction to Stochastic Processes in Biostatistics. New York: John Wiley & Sons.

11.

Dahlin, D. C. 1978. Bone Tumors, 3rd ed. Springfield, Ill.: Charles C Thomas.

12.

Davis, W. B. 1914. Development and Anatomy of the Nasal Accessory Sinuses in Man. Philadelphia: W. B. Saunders.

13.

Drettner, B. 1982. The paranasal sinuses. Pp. 145–162 in The Nose: Upper Airway Physiology and the Atmospheric Environment, D. F. Proctor, editor; and I. Andersen, editor. , eds. Amsterdam: Elsevier Biomedical Press.

14.

Evans, R. D. 1933. Radium poisoning; a review of present knowledge. Am. J. Public Health 23:1017–1023. [PMC free article: PMC1558329] [PubMed: 18013838]

15.

Evans, R. D. 1966. The effect of skeletally deposited alpha-ray emitters in man. Br. J. Radiol. 39:881–895. [PubMed: 5225013]

16.

Evans, R. D., A. T. Keane, R. J. Kolenkow, W. R. Neal, and M. M. Shanahan. 1969. Radiogenic tumors in the radium and mesothorium cases studied at M.I.T. Pp. 157–194 in Delayed Effects of Bone-Seeking Radionuclides, C. W. Mays, editor; , W. S. S. Jee, editor; , R. D. Lloyd, editor; , B. J. Stover, editor; , J. H. Dougherty, editor; , and G. N. Taylor, editor. , eds. Salt Lake City: University of Utah Press.

17.

Evans, R. D., A. T. Keane, and M. M. Shanahan. 1972. Radiogenic effects in man of long-term skeletal alpha-irradiation. Pp. 431–468 in Radiobiology of Plutonium, B. J. Stover, editor; and W. S. S. Jee, editor. , eds. Salt Lake City: The J. W. Press.

18.

Finkel, A. J., C. E. Miller, and R. J. Hasterlik. 1969. Radium-induced malignant tumors in man. Pp. 195–225 in Delayed Effects of Bone-Seeking Radionuclides, J. Mays, editor; , R. D. Lloyd, editor; , B. J. Stover, editor; , J. H. Dougherty, editor; , and G. N. Taylor, editor. , eds. Salt Lake City: University of Utah Press.

19.

Frankel, B. 1906. Über die Beziehungen der Grossenvariationen der Highmorshohlen sum individuellen Schadelbau und deren praktische Bedeutung für die Therapie der Kieferhohleneiterungen. Arch. Laryngol. Rhinol. 18:229–257.

20.

Groer, P. G., and J. H. Marshall 1976. Hazard plotting and estimates for the tumor rate and the tumor growth time for radiogenic osteosarcomas in man. Pp. 17–21 in Radiological and Environmental Research Division Annual Report. Report No. ANL-76-88, Part II. Argonne, Ill.: Argonne National Laboratory.

21.

Harris, M. J., and R. A. Schlenker. 1981. Quantitative histology of the mucous membrane of the accessory nasal sinuses and mastoid cavities. Ann. Otol. Rhinol. Laryngol. 90:33–27. [PubMed: 7469294]

22.

Hasterlik, R. J. 1960. Radiation neoplasia. Proceedings of the Institute of Medicine, Chicago: Vol. 23 , No. 2.

23.

Hasterlik, R. J., L. J. Lawson, and A. J. Finkel. 1968. Ophthalmologic aspects of carcinoma of the sphenoid sinus induced by radium poisoning. Am. J. Ophthalmol. 66:55–58. [PubMed: 5659313]

24.

Hentzer, E. 1970. Histologic studies of the normal mucosa in the middle ear, mastoid cavities and eustachian tube. Ann. Otol. Rhinol. Laryngol. 79:825–833. [PubMed: 5526992]

25.

Hindmarsh, M., M. Owen, J. Vaughan, L. F. Lamerton, and F. W. Spiers. 1958. The relative hazards of strontium 90 and radium-226. Br. J. Radiol. 31:518. [PubMed: 13584743]

26.

Hindmarsh, M., M. Owen, and J. Vaughan. 1959. A note on the distribution of radium and a calculation of the radiation dose non-uniformity factor for radium-226 and strontium-90 in the femur of a luminous dial painter. Br. J. Radiol. 32:183–187. [PubMed: 13638538]

27.

Hoecker, F. E., and P. G. Roofe. 1949. Structural differences in bone matrix associated with metabolized radium. Radiology 52:856–864. [PubMed: 18151831]

28.

Hoecker, F. E., and P. G. Roofe. 1951. Studies of radium in human bone. Radiology 56:89–98. [PubMed: 14808585]

29.

International Commission on Radiological Protection (ICRP). 1973. Alkaline Earth Metabolism in Adult Man. ICRP Publication 20. Oxford: Pergamon. (Also in Health Phys. 24:125–221 , 1973.) [PubMed: 4688295]

30.

Kolenkow, R. J. 1967. Alpha-ray dosimetry of the bone-tissue interface with application to sinus dosimetry in the radium cases. Pp. 163–201 in Annual Progress Report No. MIT-952-4. Cambridge, Mass.: Radioactivity Center, Massachusetts Institute of Technology.

31.

Littman, M. S., I. E. Kirsh, and A. T. Keane. 1978. Radium-induced malignant tumors of the mastoids and paranasal sinuses. Am. J. Roentgenol. 131:773–785 , 1978. [PubMed: 101027]

32.

Lloyd, E. The distribution of radium in human bone. Br. J. Radiol. 34:521–528. [PubMed: 13762911]

33.

Lloyd, E., and C. B. Henning. 1983. Cells at risk for the production of bone tumors in radium exposed individuals: An electron microscope study. Health Phys. 44 (Suppl. 1):135–148. [PubMed: 6862893]

34.

Loutit, J. F. 1970. Malignancy from radium. Br. J. Cancer 24:195–207. [PMC free article: PMC2008571] [PubMed: 5271269]

35.

Lyman, G. H., C. G. Lyman, and W. Johnson. 1985. Association of leukemia with radium groundwater contamination. J. Am. Med. Assoc. 254:621–626. [PubMed: 4009896]

36.

Marshall, J. H. 1962. Radioactive hotspots, bone growth and bone cancer: Self-burial of calcium-like hotspots. Pp. 35–50 in Radioisotopes and Bone, P. LaCroix, editor; and A. M. Bundy, editor. , eds. Oxford, England: Blackwell Scientific Publications. [PubMed: 15445454]

37.

Marshall, J. H. 1969. Measurements and models of skeletal metabolism. Pp. 1–122 in Mineral Metabolism, Vol. III , C. L. Comar, editor; and F. Bronner, editor. , eds. New York: Academic Press.

38.

Marshall, J. H., and P. G. Groer. 1977. A theory of the induction of bone cancer by alpha radiation. Radiat. Res. 71:149–192. [PubMed: 877237]

39.

Marshall, J. H., and C. Onkelix. 1968. Radial diffusion and the power function retention of alkaline earth radioisotopes in adult bone. Nature 217:742–744. [PubMed: 5641125]

40.

Marshall, J. H., P. G. Groer, and R. A. Schlenker. 1978. Dose to endosteal cells and relative distribution factors for radium-224 and plutonium-239 compared to radium-226. Health Phys. 35:91–101. [PubMed: 721487]

41.

Martland, H. S. 1926. Microscopic changes of certain anemias due to radioactivity. Arch. Pathol. Lab. Med. 2:465–472.

42.

Martland, H. S. 1931. The occurrence of malignancy in radioactive persons. Am. J. Cancer 15:2435–2516.

43.

Martland, H. S. 1939. Occupational tumors, bones. In Encyclopedia of Health and Hygiene. Geneva: International Labor Organization.

44.

Mays, C. W., and R. D. Lloyd. 1972. Bone sarcoma incidence vs. alpha particle dose. Pp. 409–430 in Radiobiology of Plutonium, B. J. Stover, editor; and W. S. S. Jee, editor. , eds. Salt Lake City: The J. W. Press.

45.

Mays, C. W., and H. Spiess. 1983. Epidemiological studies of German patients injected with ²²⁴Ra. Pp. 159–166 in Epidemiology Applied to Health Physics. Proceedings of the Sixteenth Mid-Year Topical Meeting of the Health Physics Society. CONF-830101. Springfield, Va.: National Technical Information Service Society.

46.

Mays, C. W., and H. Spiess. 1984. Bone sarcomas in patients given radium-224. Pp. 241–252 in Radiation Carcinogenesis. Epidemiology and Biological Significance, J. B. Boice, editor; and J. F. Fraumeni, editor. , eds. New York: Raven.

47.

Mays, C. W., T. F. Dougherty, G. N. Taylor, R. D. Lloyd, B. J. Stover, W. S. S. Jee, W. R. Christensen, J. H. Dougherty, and D. R. Atherton. 1969. Radiation-induced bone cancer in beagles. Pp. 387–408 in Delayed Effects of Bone-Seeking Radionuclides, C. W. Mays, editor; , W. S. S. Jee, editor; , R. D. Lloyd, editor; , B. J. Stover, editor; , J. H. Doughtery, editor; , and G. N. Taylor, editor. , eds. Salt Lake City: University of Utah Press.

48.

Mays, C. W., H. Spiess, G. N. Taylor, R. D. Lloyd, W. S. S. Jee, S. S. McFarland, D. H. Taysum, T. W. Brammer, D. Brammer, and T. A. Pollard. 1976. Estimated risk to human bone from ²³⁹Pu. Pp. 343–362 in The Health Effects of Plutonium and Radium, W. S. S. Jee, editor. , ed. Salt Lake City: The J. W. Press.

49.

Mays, C. W., H. Spiess, and A. Gerspach. 1978. Skeletal effects following ²²⁴Ra injections into humans. Health Phys. 35:83–90. [PubMed: 281324]

50.

Mays, C. W., H. Spiess, D. Chmelevsky, and A. Kellerer. 1986. Bone sarcoma cumulative tumor rates in patients injected with ²²⁴Ra. Pp. 27–31 in The Radiobiology of Radium and Thorotrast, W. Gössner, editor. , ed. Baltimore: Urban and Schwarzenberg.

51.

Mygind, N., M. Pedersen, and M. H. Nielsen. 1982. Morphology of the upper airway epithelium. Pp. 71–97 in The Nose: Upper Airway Physiology and the Atmospheric Environment, D. F. Proctor, editor; and I. Andersen, editor. , eds. Amsterdam: Elsevier Biomedical Press.

52.

National Cancer Institute. 1981. Surveillance, Epidemiology, and End Results: Incidence and Mortality Data, 1973–1977. Monograph No. 57. NIH Publication No. 81-2330. Bethesda, Md.: National Cancer Institute. [PubMed: 7278951]

53.

National Cancer Institute. 1982. Cancer Mortality in the United States: 1950–1977. Monograph No. 59. NIH Publication No. 82-2435. Bethesda, Md.: National Cancer Institute.

54.

National Research Council, Committee on the Biological Effects of Ionizing Radiations (BEIR). 1980. The EfFects on Populations of Exposure to Low Levels of Ionizing Radiation. Washington, D.C.: National Academy Press. 524 pp.

55.

Petersen, N. J. 1966. Midwest Environmental Health Study. Interim Report. Region V. Chicago: Environmental Protection Agency.

56.

Petersen, N. J., L. D. Samuels, H. F. Lucas, and S. P. Abrahams. 1966. An epidemiologic approach to low-level ²²⁶Ra exposure. Public Health Rep. 81:805–814. [PMC free article: PMC1919910] [PubMed: 4957940]

57.

Polednak, A. P. 1978. Bone cancer among female radium dial workers. Latency periods and incidence rates by time after exposure. J. Natl. Cancer Inst. 60:77–82 (brief communication). [PubMed: 628024]

58.

Polednak, A. P., A. F. Stehney, and R. E. Rowland. 1978. Mortality among women first employed before 1930 in the U.S. radium dial-painting industry. Am. J. Epidemol. 107:179–195. [PubMed: 629255]

59.

Pool, R. R., J. P. Morgan, N. J. Parks, J. Farnham, J. E. Littman, and M. S. Littman. Comparative pathogenesis of radium-induced intracortical bone lesions in humans and beagles. Health Phys. 44(Suppl. 1):155–177. [PubMed: 6862896]

60.

Raabe, O. G. 1984. Comparison of the carcinogenicity of radium and bone-seeking actinides. Health Phys. 46:1241–1258. [PubMed: 6327574]

61.

Raabe, O. G., S. A. Book, and N. J. Parks. 1980. Bone cancer from radium: Canine dose response explains data for mice and humans. Science 208:61–64. [PubMed: 7361106]

62.

Raabe, O. G., S. A. Book, and N. J. Parks. 1983. Lifetime bone cancer dose-response relationships in beagles and people from skeletal burdens of ²²⁶Ra and ⁹⁰Sr. Health Phys. 44:33–48. [PubMed: 6862910]

63.

Rotblatt, J., and G. Ward. 1956. Analysis of the radioactive content of tissues by alpha-track autoradiography. Phys. Med. Biol. 1:57–70. [PubMed: 13379002]

64.

Rowland, R. E. 1975. The risk of malignancy from internally-deposited radioisotopes. Pp. 146–155 in Radiation Research, Biomedical, Chemical, and Physical Perspectives, O. F. Nygaard, editor; , H. I. Adler, editor; , and W. K. Sinclair, editor. , eds. New York: Academic Press.

65.

Rowland, R. E., and J. H. Marshall. 1959. Radium in human bone: The dose in microscopic volumes of bone. Radiat. Res. 11:299–313. [PubMed: 14439709]

66.

Rowland, R. E., A. T. Keane, and P. M. Failla. 1971. The appearance times of radium-induced malignancies. Pp. 20–22 in Radiological Physics Division Annual Report, Report No. ANL-7860, Part II. Argonne, Ill.: Argonne National Laboratory.

67.

Rowland, R. E., A. F. Stehney, A. M. Brues, M. S. Littman, A. T. Keane, B. C. Patten, and M. M. Shanahan. 1978. Current status of the study of ²²⁶Ra and ²²⁸Ra in humans at the Center for Human Radiobiology. Health Phys. 35:159–166. [PubMed: 721483]

68.

Rowland, R. E., A. F. Stehney, and H. F. Lucas, Jr. 1978. Dose-response relationships for female radium dial workers. Radiat. Res. 76:368–383. [PubMed: 287126]

69.

Rowland, R. E., A. F. Stehney, and H. F. Lucas. 1983. Dose-response relationships for radium-induced bone sarcomas. Health Phys. 44(Suppl. 1):15–31. [PubMed: 6862895]

70.

Rundo, J., A. T. Keane, and M. A. Essling. 1985. Long-term retention of radium in female former dial workers. Pp. 77–85 in Metals in Bone, N. D. Priest, editor. , ed. Lancaster, England: MTP Press.

71.

Rundo, J., A. T. Keane, H. F. Lucas, R. A. Schlenker, J. H. Stebbings, and A. F. Stehney. 1986. Current (1984) status of the study of ²²⁶ Ra and ²²⁸Ra in humans at the center for human radiobiology. Pp. 14–21 in Radiobiology of Radium and Thorotrast, W. Gössner, editor; , G. B. Gerber, editor; , U. Hagen, editor; , and A. Luz, editor. , eds. Munich, West Germany: Urban and Schwarzenberg.

72.

Schaeffer, J. P. 1920. The embryology, development and anatomy of the nose, paranasal sinuses, nasolacrimal passageways and olfactory organ in man. Philadelphia: P. Blakiston's Son and Co.

73.

Schlenker, R. A. 1980. Dosimetry of paranasal sinus and mastoid epithelia in radium-exposed humans. Pp. 1–21 in Radiological and Environmental Research Division Annual Report, No. ANL-80-115, Part II. Argonne, Ill.: Argonne National Laboratory.

74.

Schlenker, R. A. 1982. Risk estimates for bone. Pp. 153–163 in Critical Issues in Setting Radiation Dose Limits. Proceedings No. 3. Bethesda, Md.: National Council on Radiation Protection and Measurements.

75.

Schlenker, R. A. 1983. Mucosal structure and radon in head carcinoma dosimetry. Health Phys. 44:556–562. [PubMed: 6853176]

76.

Schlenker, R. A. 1985. The distribution of radium and plutonium in human bone. Pp. 127–147 in Metals in Bone, N. D. Priest, editor. , ed. Lancaster, England: MTP Press.

77.

Schlenker, R. A., and J. E. Farnham. 1976. Microscopic distribution of Ra-226 in the bones of radium cases: A comparison between diffuse and average Ra-226 concentrations. Pp. 437–449 in The Health Effect of Plutonium and Radium, W. S. S. Jee, editor. , ed. Salt Lake City: The J. W. Press.

78.

Schlenker, R. A., and J. H. Marshall. 1975. Thicknesses of the deposits of plutonium at bone surfaces in the beagle. Health Phys. 29:649–654. [PubMed: 1238380]

79.

Schlenker, R. A., and B. G. Oltman. In press. High concentrations of Ra-226 and Am-241 at human bone surfaces: Implications for the ICRP 30 Bone dosimetry model. Rad. Prot. Dosimetry.

80.

Schlenker, R. A., and J. M. Smith. 1986. Argonne-Utah studies of Ra-224 endosteal surface dosimetry. Pp. 93–98 in The Radiobiology of Radium and Thorotrast, W. Gössner, editor; , G. B. Gerber, editor; , U. Hagen, editor; , and A. Luz, editor. , eds. Munich, West Germany: Urban and Schwarzenburg.

81.

Schumacher, G. H., H. J. Heyne, and R. Fanghnel. 1972. Zur Anatomie der menschlichen Nasennebenhohlen. Anat. Anz. 130:113–157. [PubMed: 5026253]

82.

Silbiger, H. 1950–1951. Über das ausmass der Mastoidpneumatiation beim Menschen. Acta Anat. 11:215–245. [PubMed: 14810422]

83.

Spiers, F. W., H. F. Lucas, J. Rundo, and G. A. Anast. 1983. Leukemia incidence in the U.S. dial workers. Health Phys. 44(Suppl. 1):65–72. [PubMed: 6575002]

84.

Spiess, H. 1969. ²²⁴Ra-induced tumors in children and adults. Pp. 227–247 in Delayed Effects of Bone-Seeking Radionuclides, C. W. Mays, editor; , W. S. S. Jee, editor; , R. D. Lloyd, editor; , B. J. Stover, editor; , J. H. Dougherty, editor; , and G. N. Taylor, editor. , eds. Salt Lake City: University of Utah Press.

85.

Spiess, H., and C. W. Mays. 1970. Bone cancers induced by Ra-224 (ThX) in children and adults. Health Phys. 19:713–729. [PubMed: 4935546]

86.

Spiess, H., and C. W. Mays. 1973. Protraction effect on bone sarcoma induction of ²²⁴Ra in children and adults. Pp. 437–450 in Radionuclides Carcinogenesis, CONF-720505, C. L. Sanders, editor; , R. H. Busch, editor; , J. E. Ballou, editor; , and D. D. Mahlum, editor. , eds. Springfield, Va.: National Technical Information Service.

87.

Spiess, H., H. Poppe, and H. Schoen. 1962. Strahlenindizierte Knochentumoren nach Thorium X-Behandlung. Monatsschrift für Kinderheilkunde 110:198–201. [PubMed: 13915776]

88.

Spiess, H., A. Gerspach, and C. W. Mays. 1978. Soft-tissue effects following ²²⁴Ra injections into humans. Health Phys. 35:61–81. [PubMed: 721486]

89.

Stebbings, J. H., H. F. Lucas, and A. F. Stehney. 1984. Mortality from cancers of major sites in female radium dial workers. Am. J. Ind. Med. 5:435–459. [PubMed: 6731445]

90.

Thurman, G. B., C. W. Mays, G. N. Taylor, A. T. Keane, and H. A. Sissons. 1973. Skeletal location of radiation-induced and naturally occurring osteosarcomas in man and dog. Cancer Res. 33:1604–1607. [PubMed: 4516008]

91.

Tos, M. 1982. Goblet cells and glands in the nose and paranasal sinuses. Pp. 99–144 in The Nose: Upper Airway Physiology and the Atmospheric Environment, D. F. Proctor, editor; and I. Andersen, editor. , eds. Amsterdam: Elsevier Biomedical Press.

92.

van Alyea, O. E. 1983. Ethmoid labyrinth. Arch. Otolaryngol. 29:881–902.

93.

Wick, R. R., and W. Gössner. 1983. Followup study of late effects in ²²⁴Ra-treated ankylosing spondylitis patients. Health Phys. 44(Suppl 1):187–195. [PubMed: 6862898]

94.

Wick, R. R., and W. Gössner. 1983. Incidence of tumours of the skeleton in ²²⁴Ra-treated ankylosing spondylitis patients. Pp. 281–288 in Biological Effects of Low-Level Radiation. Vienna: International Atomic Energy Agency.

95.

Wick, R. R., D. Chmelevsky, and W. Gössner. 1986. ²²⁴Ra risk to bone and haematopoietic tissue in ankylosing spondylitis patients. Pp. 38–44 in The Radiobiology of Radium and Thorotrast, W. Gössner, editor; , G. B. Gerber, editor; , U. Hagen, editor; , and A. Luz, editor. , eds. Baltimore: Urban and Schwarzenberg.

96.

Wolff, D., R. J. Bellucci, and A. A. Egsston. 1957. Mircoscopic anatomy of the temporal bone. Baltimore: The Williams & Wilkins Co.