Int. J. Radiarion Oncoiogy Bid. Phys.. Vol. 18, pp. 993-KQO Printal in the U.S.A. All rights reserved.
0360-3016/90 $3.00 + .W Copyright 0 1990 Pergamon Press pk
0 ASTRO Special Feature RADIATION-INDUCED CANCER AS A FACTOR IN CLINICAL DECISION (THE 1989 ASTRO GOLD MEDAL ADDRESS)
MAKING
ROBERT G. PARKER, M.D. UCLA Medical Center, Department of Radiation Oncology, 10833 LeConte Ave., Los Angeles, CA 90024-1714
INTRODUCTION
For example for C3H/ 1OT 1/2 cells, at doses less than 1.5 Gy, dose-splitting actually increases the incidence of cellular transformation, and while for doses above 1.0 Gy the dose-response curve seems to have a quadratic relationship, below 0.3 Gy the slape of the curve is unity implying a direct relationship.
It is unequivocal that cancers may be induced in man by ionizing radiations (46). It is paradoxical that this agent both cures and causes cancer (1). In clinical medicine. this potential carcinogenicity of a very useful therapeutic agent raises questions: What is the risk to a patient in a specific situation? Is such a risk acceptable, particularly in relation to the risks of other treatments? How should this information be presented to the patient, relatives and physicians to support rational decision-making? Identification of ionizing radiations as carcinogens in specific situations may be difficult; as noted by Mole (46). “the risks are many orders of magnitude less than the natural”. Inasmuch as there is no specific characteristic that without a doubt etiologically relates an individual cancer to ionizing radiations, identification must rest on statistical and sometimes circumstantial evidente. METHODS
Clinical assessment of risks An assessment of the risks of radiation-induced cancer for patients who already have cancer, necessitates a perspective different from that of an assessment of the risks for the genera1 population. Analysis can be divided into factors related to the ionizing radiations, the underlying biology of cancer and the host. Radiation-related factors Total radiation dose. Actual risk assessments are more complicated than an assumption of a linear relationship to total dose without a threshold. For low energy transfer (LET) radiations, the malignant transformation of cells appears proportional to the total dose squared until cellkilling significantly reduces the number of cells at risk. Thus, when the probability of induction of the process is low and the probability of progenitor cel1 killing increases, the risk of radiation-induced cancer per unit dose increases to a maximum and then decreases as the total dose increases (46). This changing risk per unit of radiation dose, as the total dose increases, varies with different mammalian cancers (58). The higher risk per unit dose at lower doses may explain the seemingly greater frequency of cancers reported after smal1 total doses than following high doses delivered therapeutically for cancer. For example, while cancers of the thyroid (32), and breast (5) and leukemia ( 10) have been reported following total doses of a few hundred cGy, there has been no documented increase of second cancers following doses in excess of 4500 cGy for the treatment of cancers of the cervix (30)
AND MATERIALS
Biological mechanisms There appears to be no single mechanism of carcinogenesis. The initiation of most events in tumor-induction may be similar for most tumors with the major variation being in expression ( 16). The nature of the primary lesion induced by ionizing radiations is unknown (63). Theoretically, ionizing radiations may induce cancers by direct action on target cells, or by complex actions on target cells and host tissues or they may merely advance the times of appearance of naturally occurring tumors ( 17). Gray’s hypothesis (2 l), that tumorigenesis is the net effect of a low probability inductive process and a high probability cel1 killing process, may explain why tumorigenesis may ultimately decrease with increasing doses and may increase when cel1 killing is reduced ( 14). Hall and Miller (23) noted that while fractionation and protraction nearly always reduce the biological effectiveness of a given total dose of radiations, this may not be true for tumorigenesis. Accepted for publication 16 November 1989. 993
994
1. J. Radiation Oncology 0 Biology 0 Physics
head and neck (49) or breast (50). The possibility of greater risk in those tissues immediately adjacent to the treatment volume, which receive lower doses, has not been detected. Fortunately, dose-response data for tumorigenesis in the human following cancer treatment with low LET radiations are sparse. Pattern of application. Cancers in humans have arisen following single doses, such as with the atomic bomb, following a few irregularly spaced increments, such as when infants received thymic irradiation, following daily fractionation over several weeks for the treatment of patients with cancer, and following prolonged occupational exposure in radiation workers. Inasmuch as normal tissue damage is greater for a total dose administered in a few large increments compared to delivery in many smaller increments and there may be a correlation of cancer development with severe normal tissue damage, a correlation with few large dose increments might be predicted. Sheline (personal communication, 1984) stated that nearly al1 malignant tumors arising within the volume irradiated for the suppression of pituitary adenomas followed the use of a few spaced large dose increments and repeated courses of treatment separated by several months. While in laboratory animals, fractionation of low LET radiations reduced the incidence of induced tumors ( 17) at very low doses, tumorigenesis may increase (23). In contrast, fractionation and protraction of high LET radiations may increase tumorigenesis and life shortening ( 16). Dose rate. If the radiation induction of cancer requires two or more events within a smal1 volume in a short time, then for low LET radiations, lowering the dose rate could reduce the tumorigenicity. In an animal experiment with an end point of excess ovarian tumors attributed to radiations, there was a linear dose relationship with a decreasing slope between 112 and 1.75 cGy per hr (66). However, such smal1 dose rates are rarely used in the clinic. The National Council for Radiation Protection (NCRP) (48) has estimated that low dose rates of low LET radiations may be 2-10 times less effective in tumorigenesis than are high dose rates (dose effectiveness factor). In contrast, for densely ionizing radiations (high LET), multiple events may be produced nearly simultaneously within a smal1 target volume and so dose rate reduction is unlikely to have an effect. Dose distribution. Although the risk of tumorigenesis must increase as the number of cells at risk increases, there are no specific clinical data to support this concept. When large tissue volumes are irradiated such as in TBI, TNI or the treatment of Hodgkin’s Disease, associated alterations of the immune system obscure any measurable effect. Likewise, an increased risk of cancer induction has not been associated with the inhomogeneity of dose distribution inevitable with the therapeutic interstitial implantation of radioactive isotopes. Radiation quality. The rate of transformation of mammalian cells has a quadratic dose relationship for low LET radiations and a linear dose relationship for high LET
May 1990, Volume 18, Number 5
radiations. The relative biological effectiveness (RBE) for tumorigenesis increases as the doses decrease until a first power relationship becomes operative for low LET as wel1 as high LET radiations (17). The RBE for high LET radiations for several tumors induced in laboratory animals ranges up to five (62). However, the wide variations require evaluation for specific tumor types and target organs. Biologie factors Susceptibility and tumor incidence. If there is a corre-
lation between the natura1 incidence of a cancer and susceptibility to the induction of this cancer by ionizing radiations, then the risk estimate should be relative rather than absolute (16). Although data are sparse, there is a good correlation between the frequency of induced and the frequency of naturally occurring tumors in several experimental animals. Fry ( 16) noted that when the excess risk of a radiation-induced cancer is expressed as a percentage of the natura1 incidence of a specific tumor, the risks for mouse and man are similar. He estimated that the percentage increase per cGy is 0.05-0.024. One of the difficulties in studying radiation carcinogenesis is that ionizing radiations are weak carcinogens (Dol], R. personal communication, 1980) and consequently induced cancers are rare. In an exhaustive review of reports of patients irradiated therapeutically to doses above 3,000 cGy between 1929 and 1973 at the American Oncologie Hospital, Seydel (57) found 12 carcinomas of the large bowel, 63 sarcomas of bone, 4 extraskeletal osteosarcomas, 12 malignant tumors of the skin and soft tissues, 3 thyroid cancers in patients treated for medulloblastomas, 5 carcinomas of the hypopharynx in patients treated for cancers of the larynx, and 54 patients who developed leukemias after treatment for Hodgkin’s disease, most of whom also had chemotherapy. Phillips and Sheline (52) found only two osteosarcomas at 5 and 11 years after therapeutic irradiation inclusive of radiographically normal bone, an incidence of 0.1% of the 2300 5-year survivors and 0.03% of the total of 6000 patients irradiated. In a review of several thousand patients treated with supervoltage X rays over a 20-year period at the Tumor Institute of the Swedish Hospital, Seattle, there were 4 recta1 cancers in over 1000 patients treated for cancer of the cervix, an osteosarcoma in the clavicle of a patient cured of cancer of the cervical esophagus, and one patient who developed leukemia shortly after treatment for seminoma (unpublished review, R. G. Parker, 1958). Hutchinson (3 1) found no excess of leukemia in patients irradiated for cancers of the cervix in a study with 60,000 person-years observation. Tumor types. Nearly al1 types of cancers arising in al1 tissues have been attributed to previous irradiation (30). Sarcomas are the most frequent tumors reported probably related to the inclusion of connective tissue in al1irradiated volumes. Solid tumors have been reported more frequently than leukemias. In patients receiving smal1 total doses of X rays for anklylosing spondylitis, the ratio was
Radiation-induced cancer as a factor in clinical decision making 0 R. G. PARKER
5 to 1 (45). These data for low LET radiations can not be extrapolated to fast neutrons used therapeutically because the specific RBE for carcinogenesis has not been established. For heavy ion radiations (above 500 keV/pm), the risk may be comparable to that of low LET radiations (28). Fission neutrons may have a high LET for leukemogenesis (56). Degree of malignanq. Alterations of the aggressiveness of various cancers has been attributed to therapeutic interventions, both by surgery and irradiation. Although this dubious correlation is likely to be related to the time of therapeutic intervention in the natura1 course of the cancer, in laboratory animals, the degree of malignancy, measured by lethality, has been increased by ionizing radiations (18). Although the increase in lethal Harderian gland tumors in mice was smal1 following irradiation with gamma rays, particularly with low dose rates and fractionation, there was a dose dependent increase for neutrons, which was increased by fractionation (19). Latent period. A prolonged post irradiation interval is a prerequisite for the diagnosis of radiation-induced cancers. This is necessary to separate new tumors from tumors which persist after treatment, particularly if the histology is similar. Characteristically long latent periods for solid tumors, such as more than 15 years for thyroid cancers (26) 25 years for skin cancers (53) and 24 years for head and neck cancers (34) are longer than reported for the development of leukemias. These long latent periods for most cancers do not rule out the possibility of occasional induced cancers arising shortly after irradiation. Inasmuch as a majority of patients treated for cancer are older than 50 years (17) the practica1 threat of radiation-induced cancer is less than the absolute biological risk. Such a practica1 age-related reduction of risk is not operational in the treatment of children and so should be considered separately. Multiplicity of cancers. It is wel1 established that many patients with one cancer (i.e. those arising in the head and neck, breast, retina, skin, bowel) have a higher risk of additional cancers than does the genera1 population. Thus, as noted by Mole (46) it is difficult to detect a smal1 increase of cancers when there is a high background frequency of naturally occurring cancers. Host factors Age. The strongest determinant of the risk for the development of cancer in mammals is age (16). The incidence of most cancers increases with increasing age. The incidence, tumor type, latent period and clinical consequence of tumors attributed to ionizing radiations are directly related to the age of the host at the time of irradiation. For cancers arising in the thyroid, the young beyond infancy seem most vulnerable (16). For those irradiated for ankylosing spondylitis, the susceptibility to develop leukemia increased with age (11). For survivors of the atomic bomb, the young and old appeared more susceptible to leukemia than did those of middle age (47).
995
Sex. Cancers attributed to ionizing radiations seem to reflect the relative frequency and type of tumors seen in males and females. For example, in our study of second primary cancers arising in the head and neck (49) 23 of 39 (59%) were in males and no type of secondary cancer was seen exclusively in patients of either sex. Genotype. The host% genotype influences the susceptibility to cancer (18). Many, and perhaps all, human cancers are initiated by somatic mutations (35). To-date, two classes of such genes have been identified: cellular oncogenes, which are activated by the promoting portion of retroviruses; and recessive cancer genes, identified through the study of hereditary cancers (35). Although the pattern of inheritance is usually dominant, a few cancers may be recessively inherited (35). The two most frequent hereditary conditions prediposing to cancer are neurofibromatosis and polyposis coli. Clinically recognized manifestations of recessive breakage and repair disorders include xeroderma pigmentosum, ataxia telangiectasia, Fanconi’s-anemia and Bloom’s syndrome (35). Immunodeficiency. In patients with congenital immunodeficiency states, such as ataxia telangiectasia and Wiskott-Ahich syndrome, secondary lymphomas, usually of B-lymphocyte origin, often develop (54). There is a high risk of second cancers in long-term survivors of renal and cardiac transplants (5 1). Again, most of these are diffuse lymphomas of B-lymphocyte origin. Inasmuch as T-lymphocyte function is abnormal in patients with Hodgkin’s disease (33) and is further impaired by therapy (29) this could be an underlying mechanism for the development of secondary B-cel1 lymphomas, which commonly present in the abdomen (9). Specijïc tissues involved. If the development of radiation-induced cancers is related to the total number of susceptible target cells, then connective tissues, which are present in nearly al1 therapeutically irradiated volumes, should be the origin of many secondary cancers and indeed sarcomas are the most frequently reported secondary cancers attributed to low LET radiation therapy (17). However, there has been no increase above the expected incidence of tumors arising in connective tissue, skin or vasculature in atomic bomb survivors (53). Exposnre to other carcinogens. Inasmuch as ionizing radiations are weak carcinogens (Doll, R. personal communication, 1980) exposure to other carcinogens is a major factor in tumor induction. Exposure to most non-iatrogenic carcinogens is not likely to be recognized. Drugs, used therapeutically, have long been recognized as carcinogens. The use of stilbestrol in pregnant women has been followed by the development of adenocarcinomas of the vaginas in their daughters (27). Malignant lymphomas have been reported in patients with lymphoid reactions to Dilantin (2). Skin cancers have developed after prolonged ingestion of arsenic (43). The frequency of acute leukemias increased in patients with multiple myeloma treated with melphalan or cyclophosphamide (36).
996
1. J. Radiation Oncology 0 Biology 0 Physics
However, the greatest current interest is in those patients receiving chemotherapy, with or without ionizing radiations, for the control of Hodgkin’s disease. In a study at Stanford, where the drug regimen was either MOPP (nitrogen mustard, vincristine, procarbazine, + predisone) or PAVe (procarbazine, alkeran or melphalan, vincristine), the number of second cancers was twice as high in those patients receiving adjuvant chemotherapy as in those only irradiated and this increase was entirely due to leukemia (9). None of 448 patients receiving x-ray therapy alone developed leukemia. Those patients developing solid tumors after radiation therapy only were older (median age-52 years) than those receiving combined chemotherapy and irradiation (median age-23 years). In a report by D’Angio and associates (12), the expected frequency of second tumors attributed to radiation therapy was decreased in those patients who also received actmomycin D. SpeciJic clinical situations The potential of radiation induction of cancer has been introduced into the selection of treatment for many patients with cancer. A few clinical situations wil1 be used as examples of the current status of clinical information: radiation therapy in the conservative management of the patient with breast cancer; Hodgkin’s disease; and cancer arising in the head and neck. Radiation-induced breast cancer. A major concern about the use of radiation therapy, either for breast conservation or post-mastectomy, is the risk of inducing cancer in the contralateral breast. It has been estimated that the contralateral breast absorbs 5- 10% of the maximum dose delivered to the ipsilateral breast or chest wal1 (15). The relative risk of a cancer developing in the contralateral breast has been reported to be in the range of 3.0 (24) to 5.8 (7). However, much of the data used to estimate the risk of radiation carcinogenesis has been derived from epidemiological studies of victims of the atomic bomb (60), reports of repeated fluoroscopic examinations of patients with tuberculosis, (4) and patients treated for cystic changes in the breast (3). As noted by Levitt and Mandel (40), several assumptions in these studies are open to challenge. It has been assumed that ionizing radiations are uniformly carcinogenic at any dose rate, that low and high total doses are equally carcinogenic per rad and that radiation quality, pattern of application, dose increment size and a range of host-related factors, such as age at the time of exposure, do not alter the risk. Al1 major reports have serious methodological flaws, such as absente of appropriate comparison groups, greater surveillance of the irradiated patients, over-interpretation of relative risks and inability to separate the risks related to the original cancer and the consequent irradiation (40). Consequently, Levitt and Mandel (40) concluded that the risk per rad is overestimated and even if the highest risk of a ó-fold excess for a dose of 1-400 cGy is assumed, among 20,000 patients exposed to radiation therapy less than one additional cancer would result in a lO-year observation period.
May 1990, Volume 18, Number 5
In a clinical setting, Lavey et al. (37) reviewed al1 patients treated for non-metastatic breast cancer at Duke University Medical Center between 1970 and 198 1. For 407 patients treated with surgery only, 24 developed cancer in the contralateral breast for an actuarial risk of 8.7 and a relative risk of 4.2. For the 226 patients treated by surgery and cytotoxic drugs postoperatively, 3 developed cancers in the contralateral breasts for an actuarial risk of 1.7 and a relative risk of 1.1. For 140 patients treated by surgery and postoperative irradiation, 4 developed cancers in the contralateral breasts for an actuarial risk of 3.3 and a relative risk of 2.1. For the 308 patients treated by surgery, cytotoxic agents and radiations, only 1 developed a second cancer for an actuarial risk of 1.4 and a relative risk of 0.5. These data support the position that, at least for a decade following the treatment of a breast cancer, the inclusion of chemotherapy and/or irradiation of the breast or chest wal1 does not increase the risk of developing cancer in the contralateral breast. In a study of 1630 women who were observed for 5 to 30 years after their first breast cancers were diagnosed and treated at UCLA between 1955 and 1979, there was no increase in cancers developing in the contralateral breasts for patients treated with ionizing radiations compared to those patients treated only by surgery (50). For the 32 cancers developing in the contralateral breasts of 648 patients treated only by surgery and the 16 contralateral breast cancers developing in the 836 patients receiving radiation therapy, the post treatment intervals until diagnosis of the second cancers were similar. Therefore, although there may be epidemological data that imply that undocumented low doses of ionizing radiations increased the number of breast cancers in certain populations, the risk of developing cancers in the contralateral breasts of patients irradiated for their initial breast cancers does not seem greater than the expected natura1 incidence. Second cancers related to the treatment ofpatients with Hodgkin ‘s disease. Both acute leukemias and solid tumors have developed more frequently than expected in patients treated for Hodgkin’s disease compared to standard populations. Although there have been suggestions that longterm survivors of Hodgkin’s disease had an elevated incidence of second cancers even before the current extensive use of radiation therapy and chemotherapy (55) there is a legitimate concern about the carcinogenicity of modern treatment particularly related to the use of alkylating agents and combined chemotherapy and radiation therapy. In a study of 1,507 patients treated at Stanford between 1968 and 1985 (61), the mean actuarial risk at 15 years for al1 second cancers was 17.6 + 3.1%. The largest component was a risk of 13.2 + 3.1% due to solid tumors. The risk of developing leukemia reached a plateau of 3.3 f 0.6% at 10 years. The risk of developing non-Hodgkin’s lymphoma slowly increased to 1.6 + 0.7% at the end of the 15-year follow-up period. Al1 leukemias were fatal. However, 45% of those patients developing solid tumors
Radiation-induced cancer as a factor in clinical decision making 0 R.G. PARKER
were alive at 1.6 years after diagnosis and 55% of these patients developing NHL were alive at 2.1 years. The risk of developing leukemia was less in patients receiving salvage chemotherapy than in those receiving primary chemotherapy. This may reflect a smaller total dose, because of tolerante limitations, in the former group. Although the risk seemed to increase with advancing age, the relative risk was the same for al1 ages. NO leukemias developed in patients treated only with the PAVe regimen. In response to the question whether treatment combining irradiation and cytoxic drugs was being withheld because of a concern about the risk of inducing second malignancies, Lavey, et al. (38) studied 3 13 patients with Hodgkin? disease and 686 patients with NHL treated at Duke from 1970 through 198 1. They found that the risk of developing acute leukemia had not been increased by radiation therapy alone and that the increased risk following combined radiation therapy and chemotherapy had been no greater than that following chemotherapy alone. In an extensive review of the literature, only 4 of 3,428 patients with Hodgkin’s disease receiving only radiation therapy developed acute non-lympocytic leukemia. Of those 2 16 patients developing acute leukemias after chemotherapy, 2 10 received alkylating agents (22). The most frequent second solid cancer in al1 reports is cancer of the lung. In the Stanford study (6 l), the relative risk was 7.7. Al1 second lung cancers developed in smokers. In the report of Lavey et al. (38) of 13 second solid cancers. 5 arose in the lung, 3 in the breast, and 1 each in the brain, prostate, rectum, skin and stomach. The relative risk for the development of second solid cancers had been elevated in al1 treatment groups. The relatively greater incidence in the radiation therapy group in this study as wel1 as the lack of risk for the chemotherapy only group reported by Boivin and O’Brien (6) are likely the result of methodological limitations. Second cancers in the head and neck. The detection of radiation-induced cancers in the head and neck is particularly difficult because of the frequency of naturally occurring multiple cancers. Nevertheless, it has been suggested that the threat of radiation-carcinogenesis makes surgery preferable, whenever it is applicable. Lawson and Som (39) stated that radiation therapy resulted in a 10% induction of second primary cancers in the larynx. Two postcricoid carcinomas have been reported in patients cured of laryngeal carcinomas 7 and 17 years previously by irradiation (59). Kiguchi et af. (34) and Glanz (20) reported a total of 11 patients who developed laryngeal cancers after previous irradiation. Modan et al (44) reported an increase in brain tumors and leukemias after radiation therapy for tinea capitis in children. In 1950, Duffy and Fitzgerald ( 13) first suggested an etological relationship between extemal irradiation and the development of thyroid cancers in children. Usually the latent periods for second cancers arising in the head and neck have been 10-25 years and the doses have been low (200-600 cGy). However, a few thyroid cancers have been reported following higher therapeutic
997
doses (42, 57, 65). The development of thyroid cancers in the young has led to a suggestion that radiation therapy is contraindicated for patients under 40 years of age (25). However, in a review of 58 survivors followed for 10-24 years from 272 patients irradiated when they were less than 30 years of age, no excess cancers have been detected (25). Seydel(57) reported that 8 secondary head and neck cancers developed in 407 patients (2%) treated surgically for primary head and neck cancer, whereas 23 secondary cancers developed both inside and outside the irradiated volume of 1464 patients ( 1.6%) with primary head and neck cancers. For those patients surviving more than 5 years after irradiation, the incidence was 1.5% (9/6 11). In a study of 2,15 1 patients with cancers of the head and neck diagnosed and treated at UCLA between 1955 and 1979 (49) second primary cancers arose in the head and neck in excess of 2.5 per 1000 person years at risk. The risk of developing a second cancer in the head and neck has been as great in patients treated only by surgery as in those patients receiving radiation therapy and the post treatment intervals have been similar in both groups. Risks in perspective As noted by Fry (16) “man lives with many risks and even the cure of many human diseases canies with it a risk”. The risk of second cancers in patients irradiated for initial cancers is small, but very real, and can be predicted for specific populations, although not for individuals. The latent period of several years prior to the appearance of radiation-induced cancers reflects the success of treatment of the initial cancers. Consequently, these patients have enjoyed many years of life after the control of their irradiated cancers. For many patients, the likely latent period prior to the appearance of a second cancer is longer than the life expectancy related to other threats. Also, it is often
Table 1. Operative
mortality
1967-77 (Ziffren, % mortality
1979) (67)
by age
Operation
0360-3016/90 $3.00 + .W Copyright 0 1990 Pergamon Press pk
0 ASTRO Special Feature RADIATION-INDUCED CANCER AS A FACTOR IN CLINICAL DECISION (THE 1989 ASTRO GOLD MEDAL ADDRESS)
MAKING
ROBERT G. PARKER, M.D. UCLA Medical Center, Department of Radiation Oncology, 10833 LeConte Ave., Los Angeles, CA 90024-1714
INTRODUCTION
For example for C3H/ 1OT 1/2 cells, at doses less than 1.5 Gy, dose-splitting actually increases the incidence of cellular transformation, and while for doses above 1.0 Gy the dose-response curve seems to have a quadratic relationship, below 0.3 Gy the slape of the curve is unity implying a direct relationship.
It is unequivocal that cancers may be induced in man by ionizing radiations (46). It is paradoxical that this agent both cures and causes cancer (1). In clinical medicine. this potential carcinogenicity of a very useful therapeutic agent raises questions: What is the risk to a patient in a specific situation? Is such a risk acceptable, particularly in relation to the risks of other treatments? How should this information be presented to the patient, relatives and physicians to support rational decision-making? Identification of ionizing radiations as carcinogens in specific situations may be difficult; as noted by Mole (46). “the risks are many orders of magnitude less than the natural”. Inasmuch as there is no specific characteristic that without a doubt etiologically relates an individual cancer to ionizing radiations, identification must rest on statistical and sometimes circumstantial evidente. METHODS
Clinical assessment of risks An assessment of the risks of radiation-induced cancer for patients who already have cancer, necessitates a perspective different from that of an assessment of the risks for the genera1 population. Analysis can be divided into factors related to the ionizing radiations, the underlying biology of cancer and the host. Radiation-related factors Total radiation dose. Actual risk assessments are more complicated than an assumption of a linear relationship to total dose without a threshold. For low energy transfer (LET) radiations, the malignant transformation of cells appears proportional to the total dose squared until cellkilling significantly reduces the number of cells at risk. Thus, when the probability of induction of the process is low and the probability of progenitor cel1 killing increases, the risk of radiation-induced cancer per unit dose increases to a maximum and then decreases as the total dose increases (46). This changing risk per unit of radiation dose, as the total dose increases, varies with different mammalian cancers (58). The higher risk per unit dose at lower doses may explain the seemingly greater frequency of cancers reported after smal1 total doses than following high doses delivered therapeutically for cancer. For example, while cancers of the thyroid (32), and breast (5) and leukemia ( 10) have been reported following total doses of a few hundred cGy, there has been no documented increase of second cancers following doses in excess of 4500 cGy for the treatment of cancers of the cervix (30)
AND MATERIALS
Biological mechanisms There appears to be no single mechanism of carcinogenesis. The initiation of most events in tumor-induction may be similar for most tumors with the major variation being in expression ( 16). The nature of the primary lesion induced by ionizing radiations is unknown (63). Theoretically, ionizing radiations may induce cancers by direct action on target cells, or by complex actions on target cells and host tissues or they may merely advance the times of appearance of naturally occurring tumors ( 17). Gray’s hypothesis (2 l), that tumorigenesis is the net effect of a low probability inductive process and a high probability cel1 killing process, may explain why tumorigenesis may ultimately decrease with increasing doses and may increase when cel1 killing is reduced ( 14). Hall and Miller (23) noted that while fractionation and protraction nearly always reduce the biological effectiveness of a given total dose of radiations, this may not be true for tumorigenesis. Accepted for publication 16 November 1989. 993
994
1. J. Radiation Oncology 0 Biology 0 Physics
head and neck (49) or breast (50). The possibility of greater risk in those tissues immediately adjacent to the treatment volume, which receive lower doses, has not been detected. Fortunately, dose-response data for tumorigenesis in the human following cancer treatment with low LET radiations are sparse. Pattern of application. Cancers in humans have arisen following single doses, such as with the atomic bomb, following a few irregularly spaced increments, such as when infants received thymic irradiation, following daily fractionation over several weeks for the treatment of patients with cancer, and following prolonged occupational exposure in radiation workers. Inasmuch as normal tissue damage is greater for a total dose administered in a few large increments compared to delivery in many smaller increments and there may be a correlation of cancer development with severe normal tissue damage, a correlation with few large dose increments might be predicted. Sheline (personal communication, 1984) stated that nearly al1 malignant tumors arising within the volume irradiated for the suppression of pituitary adenomas followed the use of a few spaced large dose increments and repeated courses of treatment separated by several months. While in laboratory animals, fractionation of low LET radiations reduced the incidence of induced tumors ( 17) at very low doses, tumorigenesis may increase (23). In contrast, fractionation and protraction of high LET radiations may increase tumorigenesis and life shortening ( 16). Dose rate. If the radiation induction of cancer requires two or more events within a smal1 volume in a short time, then for low LET radiations, lowering the dose rate could reduce the tumorigenicity. In an animal experiment with an end point of excess ovarian tumors attributed to radiations, there was a linear dose relationship with a decreasing slope between 112 and 1.75 cGy per hr (66). However, such smal1 dose rates are rarely used in the clinic. The National Council for Radiation Protection (NCRP) (48) has estimated that low dose rates of low LET radiations may be 2-10 times less effective in tumorigenesis than are high dose rates (dose effectiveness factor). In contrast, for densely ionizing radiations (high LET), multiple events may be produced nearly simultaneously within a smal1 target volume and so dose rate reduction is unlikely to have an effect. Dose distribution. Although the risk of tumorigenesis must increase as the number of cells at risk increases, there are no specific clinical data to support this concept. When large tissue volumes are irradiated such as in TBI, TNI or the treatment of Hodgkin’s Disease, associated alterations of the immune system obscure any measurable effect. Likewise, an increased risk of cancer induction has not been associated with the inhomogeneity of dose distribution inevitable with the therapeutic interstitial implantation of radioactive isotopes. Radiation quality. The rate of transformation of mammalian cells has a quadratic dose relationship for low LET radiations and a linear dose relationship for high LET
May 1990, Volume 18, Number 5
radiations. The relative biological effectiveness (RBE) for tumorigenesis increases as the doses decrease until a first power relationship becomes operative for low LET as wel1 as high LET radiations (17). The RBE for high LET radiations for several tumors induced in laboratory animals ranges up to five (62). However, the wide variations require evaluation for specific tumor types and target organs. Biologie factors Susceptibility and tumor incidence. If there is a corre-
lation between the natura1 incidence of a cancer and susceptibility to the induction of this cancer by ionizing radiations, then the risk estimate should be relative rather than absolute (16). Although data are sparse, there is a good correlation between the frequency of induced and the frequency of naturally occurring tumors in several experimental animals. Fry ( 16) noted that when the excess risk of a radiation-induced cancer is expressed as a percentage of the natura1 incidence of a specific tumor, the risks for mouse and man are similar. He estimated that the percentage increase per cGy is 0.05-0.024. One of the difficulties in studying radiation carcinogenesis is that ionizing radiations are weak carcinogens (Dol], R. personal communication, 1980) and consequently induced cancers are rare. In an exhaustive review of reports of patients irradiated therapeutically to doses above 3,000 cGy between 1929 and 1973 at the American Oncologie Hospital, Seydel (57) found 12 carcinomas of the large bowel, 63 sarcomas of bone, 4 extraskeletal osteosarcomas, 12 malignant tumors of the skin and soft tissues, 3 thyroid cancers in patients treated for medulloblastomas, 5 carcinomas of the hypopharynx in patients treated for cancers of the larynx, and 54 patients who developed leukemias after treatment for Hodgkin’s disease, most of whom also had chemotherapy. Phillips and Sheline (52) found only two osteosarcomas at 5 and 11 years after therapeutic irradiation inclusive of radiographically normal bone, an incidence of 0.1% of the 2300 5-year survivors and 0.03% of the total of 6000 patients irradiated. In a review of several thousand patients treated with supervoltage X rays over a 20-year period at the Tumor Institute of the Swedish Hospital, Seattle, there were 4 recta1 cancers in over 1000 patients treated for cancer of the cervix, an osteosarcoma in the clavicle of a patient cured of cancer of the cervical esophagus, and one patient who developed leukemia shortly after treatment for seminoma (unpublished review, R. G. Parker, 1958). Hutchinson (3 1) found no excess of leukemia in patients irradiated for cancers of the cervix in a study with 60,000 person-years observation. Tumor types. Nearly al1 types of cancers arising in al1 tissues have been attributed to previous irradiation (30). Sarcomas are the most frequent tumors reported probably related to the inclusion of connective tissue in al1irradiated volumes. Solid tumors have been reported more frequently than leukemias. In patients receiving smal1 total doses of X rays for anklylosing spondylitis, the ratio was
Radiation-induced cancer as a factor in clinical decision making 0 R. G. PARKER
5 to 1 (45). These data for low LET radiations can not be extrapolated to fast neutrons used therapeutically because the specific RBE for carcinogenesis has not been established. For heavy ion radiations (above 500 keV/pm), the risk may be comparable to that of low LET radiations (28). Fission neutrons may have a high LET for leukemogenesis (56). Degree of malignanq. Alterations of the aggressiveness of various cancers has been attributed to therapeutic interventions, both by surgery and irradiation. Although this dubious correlation is likely to be related to the time of therapeutic intervention in the natura1 course of the cancer, in laboratory animals, the degree of malignancy, measured by lethality, has been increased by ionizing radiations (18). Although the increase in lethal Harderian gland tumors in mice was smal1 following irradiation with gamma rays, particularly with low dose rates and fractionation, there was a dose dependent increase for neutrons, which was increased by fractionation (19). Latent period. A prolonged post irradiation interval is a prerequisite for the diagnosis of radiation-induced cancers. This is necessary to separate new tumors from tumors which persist after treatment, particularly if the histology is similar. Characteristically long latent periods for solid tumors, such as more than 15 years for thyroid cancers (26) 25 years for skin cancers (53) and 24 years for head and neck cancers (34) are longer than reported for the development of leukemias. These long latent periods for most cancers do not rule out the possibility of occasional induced cancers arising shortly after irradiation. Inasmuch as a majority of patients treated for cancer are older than 50 years (17) the practica1 threat of radiation-induced cancer is less than the absolute biological risk. Such a practica1 age-related reduction of risk is not operational in the treatment of children and so should be considered separately. Multiplicity of cancers. It is wel1 established that many patients with one cancer (i.e. those arising in the head and neck, breast, retina, skin, bowel) have a higher risk of additional cancers than does the genera1 population. Thus, as noted by Mole (46) it is difficult to detect a smal1 increase of cancers when there is a high background frequency of naturally occurring cancers. Host factors Age. The strongest determinant of the risk for the development of cancer in mammals is age (16). The incidence of most cancers increases with increasing age. The incidence, tumor type, latent period and clinical consequence of tumors attributed to ionizing radiations are directly related to the age of the host at the time of irradiation. For cancers arising in the thyroid, the young beyond infancy seem most vulnerable (16). For those irradiated for ankylosing spondylitis, the susceptibility to develop leukemia increased with age (11). For survivors of the atomic bomb, the young and old appeared more susceptible to leukemia than did those of middle age (47).
995
Sex. Cancers attributed to ionizing radiations seem to reflect the relative frequency and type of tumors seen in males and females. For example, in our study of second primary cancers arising in the head and neck (49) 23 of 39 (59%) were in males and no type of secondary cancer was seen exclusively in patients of either sex. Genotype. The host% genotype influences the susceptibility to cancer (18). Many, and perhaps all, human cancers are initiated by somatic mutations (35). To-date, two classes of such genes have been identified: cellular oncogenes, which are activated by the promoting portion of retroviruses; and recessive cancer genes, identified through the study of hereditary cancers (35). Although the pattern of inheritance is usually dominant, a few cancers may be recessively inherited (35). The two most frequent hereditary conditions prediposing to cancer are neurofibromatosis and polyposis coli. Clinically recognized manifestations of recessive breakage and repair disorders include xeroderma pigmentosum, ataxia telangiectasia, Fanconi’s-anemia and Bloom’s syndrome (35). Immunodeficiency. In patients with congenital immunodeficiency states, such as ataxia telangiectasia and Wiskott-Ahich syndrome, secondary lymphomas, usually of B-lymphocyte origin, often develop (54). There is a high risk of second cancers in long-term survivors of renal and cardiac transplants (5 1). Again, most of these are diffuse lymphomas of B-lymphocyte origin. Inasmuch as T-lymphocyte function is abnormal in patients with Hodgkin’s disease (33) and is further impaired by therapy (29) this could be an underlying mechanism for the development of secondary B-cel1 lymphomas, which commonly present in the abdomen (9). Specijïc tissues involved. If the development of radiation-induced cancers is related to the total number of susceptible target cells, then connective tissues, which are present in nearly al1 therapeutically irradiated volumes, should be the origin of many secondary cancers and indeed sarcomas are the most frequently reported secondary cancers attributed to low LET radiation therapy (17). However, there has been no increase above the expected incidence of tumors arising in connective tissue, skin or vasculature in atomic bomb survivors (53). Exposnre to other carcinogens. Inasmuch as ionizing radiations are weak carcinogens (Doll, R. personal communication, 1980) exposure to other carcinogens is a major factor in tumor induction. Exposure to most non-iatrogenic carcinogens is not likely to be recognized. Drugs, used therapeutically, have long been recognized as carcinogens. The use of stilbestrol in pregnant women has been followed by the development of adenocarcinomas of the vaginas in their daughters (27). Malignant lymphomas have been reported in patients with lymphoid reactions to Dilantin (2). Skin cancers have developed after prolonged ingestion of arsenic (43). The frequency of acute leukemias increased in patients with multiple myeloma treated with melphalan or cyclophosphamide (36).
996
1. J. Radiation Oncology 0 Biology 0 Physics
However, the greatest current interest is in those patients receiving chemotherapy, with or without ionizing radiations, for the control of Hodgkin’s disease. In a study at Stanford, where the drug regimen was either MOPP (nitrogen mustard, vincristine, procarbazine, + predisone) or PAVe (procarbazine, alkeran or melphalan, vincristine), the number of second cancers was twice as high in those patients receiving adjuvant chemotherapy as in those only irradiated and this increase was entirely due to leukemia (9). None of 448 patients receiving x-ray therapy alone developed leukemia. Those patients developing solid tumors after radiation therapy only were older (median age-52 years) than those receiving combined chemotherapy and irradiation (median age-23 years). In a report by D’Angio and associates (12), the expected frequency of second tumors attributed to radiation therapy was decreased in those patients who also received actmomycin D. SpeciJic clinical situations The potential of radiation induction of cancer has been introduced into the selection of treatment for many patients with cancer. A few clinical situations wil1 be used as examples of the current status of clinical information: radiation therapy in the conservative management of the patient with breast cancer; Hodgkin’s disease; and cancer arising in the head and neck. Radiation-induced breast cancer. A major concern about the use of radiation therapy, either for breast conservation or post-mastectomy, is the risk of inducing cancer in the contralateral breast. It has been estimated that the contralateral breast absorbs 5- 10% of the maximum dose delivered to the ipsilateral breast or chest wal1 (15). The relative risk of a cancer developing in the contralateral breast has been reported to be in the range of 3.0 (24) to 5.8 (7). However, much of the data used to estimate the risk of radiation carcinogenesis has been derived from epidemiological studies of victims of the atomic bomb (60), reports of repeated fluoroscopic examinations of patients with tuberculosis, (4) and patients treated for cystic changes in the breast (3). As noted by Levitt and Mandel (40), several assumptions in these studies are open to challenge. It has been assumed that ionizing radiations are uniformly carcinogenic at any dose rate, that low and high total doses are equally carcinogenic per rad and that radiation quality, pattern of application, dose increment size and a range of host-related factors, such as age at the time of exposure, do not alter the risk. Al1 major reports have serious methodological flaws, such as absente of appropriate comparison groups, greater surveillance of the irradiated patients, over-interpretation of relative risks and inability to separate the risks related to the original cancer and the consequent irradiation (40). Consequently, Levitt and Mandel (40) concluded that the risk per rad is overestimated and even if the highest risk of a ó-fold excess for a dose of 1-400 cGy is assumed, among 20,000 patients exposed to radiation therapy less than one additional cancer would result in a lO-year observation period.
May 1990, Volume 18, Number 5
In a clinical setting, Lavey et al. (37) reviewed al1 patients treated for non-metastatic breast cancer at Duke University Medical Center between 1970 and 198 1. For 407 patients treated with surgery only, 24 developed cancer in the contralateral breast for an actuarial risk of 8.7 and a relative risk of 4.2. For the 226 patients treated by surgery and cytotoxic drugs postoperatively, 3 developed cancers in the contralateral breasts for an actuarial risk of 1.7 and a relative risk of 1.1. For 140 patients treated by surgery and postoperative irradiation, 4 developed cancers in the contralateral breasts for an actuarial risk of 3.3 and a relative risk of 2.1. For the 308 patients treated by surgery, cytotoxic agents and radiations, only 1 developed a second cancer for an actuarial risk of 1.4 and a relative risk of 0.5. These data support the position that, at least for a decade following the treatment of a breast cancer, the inclusion of chemotherapy and/or irradiation of the breast or chest wal1 does not increase the risk of developing cancer in the contralateral breast. In a study of 1630 women who were observed for 5 to 30 years after their first breast cancers were diagnosed and treated at UCLA between 1955 and 1979, there was no increase in cancers developing in the contralateral breasts for patients treated with ionizing radiations compared to those patients treated only by surgery (50). For the 32 cancers developing in the contralateral breasts of 648 patients treated only by surgery and the 16 contralateral breast cancers developing in the 836 patients receiving radiation therapy, the post treatment intervals until diagnosis of the second cancers were similar. Therefore, although there may be epidemological data that imply that undocumented low doses of ionizing radiations increased the number of breast cancers in certain populations, the risk of developing cancers in the contralateral breasts of patients irradiated for their initial breast cancers does not seem greater than the expected natura1 incidence. Second cancers related to the treatment ofpatients with Hodgkin ‘s disease. Both acute leukemias and solid tumors have developed more frequently than expected in patients treated for Hodgkin’s disease compared to standard populations. Although there have been suggestions that longterm survivors of Hodgkin’s disease had an elevated incidence of second cancers even before the current extensive use of radiation therapy and chemotherapy (55) there is a legitimate concern about the carcinogenicity of modern treatment particularly related to the use of alkylating agents and combined chemotherapy and radiation therapy. In a study of 1,507 patients treated at Stanford between 1968 and 1985 (61), the mean actuarial risk at 15 years for al1 second cancers was 17.6 + 3.1%. The largest component was a risk of 13.2 + 3.1% due to solid tumors. The risk of developing leukemia reached a plateau of 3.3 f 0.6% at 10 years. The risk of developing non-Hodgkin’s lymphoma slowly increased to 1.6 + 0.7% at the end of the 15-year follow-up period. Al1 leukemias were fatal. However, 45% of those patients developing solid tumors
Radiation-induced cancer as a factor in clinical decision making 0 R.G. PARKER
were alive at 1.6 years after diagnosis and 55% of these patients developing NHL were alive at 2.1 years. The risk of developing leukemia was less in patients receiving salvage chemotherapy than in those receiving primary chemotherapy. This may reflect a smaller total dose, because of tolerante limitations, in the former group. Although the risk seemed to increase with advancing age, the relative risk was the same for al1 ages. NO leukemias developed in patients treated only with the PAVe regimen. In response to the question whether treatment combining irradiation and cytoxic drugs was being withheld because of a concern about the risk of inducing second malignancies, Lavey, et al. (38) studied 3 13 patients with Hodgkin? disease and 686 patients with NHL treated at Duke from 1970 through 198 1. They found that the risk of developing acute leukemia had not been increased by radiation therapy alone and that the increased risk following combined radiation therapy and chemotherapy had been no greater than that following chemotherapy alone. In an extensive review of the literature, only 4 of 3,428 patients with Hodgkin’s disease receiving only radiation therapy developed acute non-lympocytic leukemia. Of those 2 16 patients developing acute leukemias after chemotherapy, 2 10 received alkylating agents (22). The most frequent second solid cancer in al1 reports is cancer of the lung. In the Stanford study (6 l), the relative risk was 7.7. Al1 second lung cancers developed in smokers. In the report of Lavey et al. (38) of 13 second solid cancers. 5 arose in the lung, 3 in the breast, and 1 each in the brain, prostate, rectum, skin and stomach. The relative risk for the development of second solid cancers had been elevated in al1 treatment groups. The relatively greater incidence in the radiation therapy group in this study as wel1 as the lack of risk for the chemotherapy only group reported by Boivin and O’Brien (6) are likely the result of methodological limitations. Second cancers in the head and neck. The detection of radiation-induced cancers in the head and neck is particularly difficult because of the frequency of naturally occurring multiple cancers. Nevertheless, it has been suggested that the threat of radiation-carcinogenesis makes surgery preferable, whenever it is applicable. Lawson and Som (39) stated that radiation therapy resulted in a 10% induction of second primary cancers in the larynx. Two postcricoid carcinomas have been reported in patients cured of laryngeal carcinomas 7 and 17 years previously by irradiation (59). Kiguchi et af. (34) and Glanz (20) reported a total of 11 patients who developed laryngeal cancers after previous irradiation. Modan et al (44) reported an increase in brain tumors and leukemias after radiation therapy for tinea capitis in children. In 1950, Duffy and Fitzgerald ( 13) first suggested an etological relationship between extemal irradiation and the development of thyroid cancers in children. Usually the latent periods for second cancers arising in the head and neck have been 10-25 years and the doses have been low (200-600 cGy). However, a few thyroid cancers have been reported following higher therapeutic
997
doses (42, 57, 65). The development of thyroid cancers in the young has led to a suggestion that radiation therapy is contraindicated for patients under 40 years of age (25). However, in a review of 58 survivors followed for 10-24 years from 272 patients irradiated when they were less than 30 years of age, no excess cancers have been detected (25). Seydel(57) reported that 8 secondary head and neck cancers developed in 407 patients (2%) treated surgically for primary head and neck cancer, whereas 23 secondary cancers developed both inside and outside the irradiated volume of 1464 patients ( 1.6%) with primary head and neck cancers. For those patients surviving more than 5 years after irradiation, the incidence was 1.5% (9/6 11). In a study of 2,15 1 patients with cancers of the head and neck diagnosed and treated at UCLA between 1955 and 1979 (49) second primary cancers arose in the head and neck in excess of 2.5 per 1000 person years at risk. The risk of developing a second cancer in the head and neck has been as great in patients treated only by surgery as in those patients receiving radiation therapy and the post treatment intervals have been similar in both groups. Risks in perspective As noted by Fry (16) “man lives with many risks and even the cure of many human diseases canies with it a risk”. The risk of second cancers in patients irradiated for initial cancers is small, but very real, and can be predicted for specific populations, although not for individuals. The latent period of several years prior to the appearance of radiation-induced cancers reflects the success of treatment of the initial cancers. Consequently, these patients have enjoyed many years of life after the control of their irradiated cancers. For many patients, the likely latent period prior to the appearance of a second cancer is longer than the life expectancy related to other threats. Also, it is often
Table 1. Operative
mortality
1967-77 (Ziffren, % mortality
1979) (67)
by age
Operation
