In modern times the dictum of Hippocrates that a physician above all should do no harm to his or her patient has been modified to a consideration of the benefit to the patient in relation to the risk of the procedure or therapy involved. Many times the risk to the patient was unknown or unsus- pected, and it was only at a later date that the risk factor became evident. Perhaps the best exam- ple of this was the administration of diethylstilbestrol to pregnant women in order to avert a threatened abortion. The benefit of such a procedure was obvious, but the risk did not become obvious until many years later, when a small percentage of the female offspring of mothers treated with this estrogenic analog during gestation developed vaginal carcinomas, usually shortly after puberty (Herbst, 1981) (Table 11.12). The incidence of clear cell adenocarcinoma
Table 11.12 Carcinogenic Risks of Chemical Agents Associated with Medical Therapy and Diagnosis
of the vagina in women exposed to diethylstilbestrol in utero is about 1 per 1000 exposed moth- ers (Melnick et al., 1987). However, other abnormalities of the genital tract, especially a condi- tion known as vaginal adenosis, occurred in about 70% of women exposed to diethylstilbestrol in utero (Johnson et al., 1979). While no carcinogenic effects were noted in male offspring exposed in utero, several anomalies of the genital tract have been described as well as possible modifica- tion of social behaviors in men exposed in utero (Mittendorf, 1995). Prenatal exposure of mice to diethylstilbestrol induces a variety of both neoplastic and developmental abnormalities in the genital tracts of both male and female offspring, some of which mimic those seen in the human (Newbold, 1995). Recently, there have been reports that women exposed to diethylstilbestrol in utero do have a moderately increased risk for developing breast cancer (Greenberg et al., 1984); this risk does not increase with time (Calle et al., 1996). Diethylstilbestrol may be metabolized to reactive forms, as shown by in vitro studies (Tsutsui et al., 1986; Gladek and Liehr, 1989).
This alteration may lead to the formation of 32P-postlabeled DNA adducts (Liehr et al., 1985), ele-vated 8-hydroxydeoxyguanosine levels in DNA in vivo (Roy et al., 1991), and DNA damage in germ cells in mice in vivo (Racine and Schmid, 1984). Diethylstilbestrol is clastogenic in a vari- ety of systems (Birnboim, 1985), such as human leukocytes, and it induces aneuploidy in vivo and in vitro (Bishun et al., 1977). Interestingly, clear-cell adenocarcinomas of the vagina induced by diethylstilbesterol show no evidence for the induction of mutations in the K-ras or H-ras protooncogenes, the Wilms’ tumor suppressor gene, the estrogen receptor gene, or the p53 tumor suppressor gene, but evidence of microsatellite instability was found in all of the neoplasms ex- amined (Boyd et al., 1996). Thus, unlike natural estrogens, this synthetic estrogen may exert its effects by altering DNA and genomic structure during the stages of initiation and/or progression.
Other synthetic estrogens, primarily utilized as oral contraceptives, have been associated with the induction of liver cell adenomas when taken chronically by premenopausal women. The relation of risk of development of liver cell adenomas to the duration of the use of oral contra- ceptives is given in Table 11.13 (Edmondson et al., 1976). While some isolated cases have been reported as indicating an association between liver cell adenomas, dysplasia, and hepatocellular carcinoma (Tao, 1992), the epidemiological evidence to date is insufficient to conclude that there is a relation between carcinoma development and exposure to synthetic estrogens in oral contraceptives. Furthermore, studies have demonstrated that, upon withdrawal of the oral contra- ceptives, regression of adenomas and particularly a related lesion, focal nodular hyperplasia, oc- curs (Steinbrecher et al., 1981; Pain et al., 1991). Furthermore, the more recent preparations of oral contraceptives contain other steroids such as progestins, which seem to moderate the ten- dency for liver adenomas to develop. The potential relation between synthetic estrogens used in oral contraceptives and the development of breast cancer has been extensively investigated. In
premenopausal women, there is some evidence for a slight increase in risk, again related to the length of use, as noted in Figure 11.11 (McGonigle and Huggins, 1991; Rookus and van Leeu- wen, 1994). In postmenopausal women—in a study combining more than 50 epidemiological investigations—it was concluded that a small increase in the relative risk of breast cancer oc- curred within the first 10 years after cessation of the use of synthetic estrogens/progestins. Sub- sequently, the risk disappeared (Calle et al., 1996). A more striking phenomenon, however, is the relationship of unopposed (no progestins) estrogens and endometrial cancer in postmenopausal women. In this instance, the so-called hormonal replacement therapy was shown some years ago to be associated with a substantial (6- to 15-fold) risk for the development of endometrial carci- noma (Antunes et al., 1979). While the concomitant administration of progestin with the estro- gen does not completely eliminate the risk (Beresford et al., 1997) of lingering effects of unopposed estrogen after administration has ceased, most carcinogenic effects of estrogens to the endometrium can be prevented by the addition of a progestin (Barrett-Connor, 1992).
More recently, on the basis of both clinical and experimental investigations, synthetic anti- estrogens have come into play both for the therapy of breast cancer and for its prevention. At present the most commonly utilized antiestrogen is tamoxifen, which has a structure not unre- lated to that of diethylstilbestrol (Chapter 3). This chemical has been shown to decrease the re- currence of breast cancer significantly after therapy. In addition, in a large clinical trial recently completed, tamoxifen decreased the risk of developing primary breast cancer in a high-risk pop- ulation (Jordan, 1993; Fisher et al., 1998). Tamoxifen exhibits estrogen-antagonist activity in the breast and the hypothalamic-pituitary axis (Mandeville and Houde, 1994). However, in most other tissues, this chemical is a moderately active estrogen. Although in a large number of stud- ies involving tamoxifen, in both therapeutic and preventive situations, little or no carcinogenic risk was demonstrated (Morgan, 1997), a significant number of investigations have indicated that tamoxifen therapy increases the risk of endometrial cancer sevenfold or more (Fisher et al.,1994; Morgan, 1997). Despite this finding, the benefit of using tamoxifen far exceeds this risk factor (Fisher et al., 1994; Assikis et al., 1996). Tamoxifen’s carcinogenic action appears to be similar to that of diethylstilbestrol in that it and its congeners induce 32P-postlabeled DNA ad- ducts (Li et al., 1997). These antiestrogens are effective promoting agents in experimental sys-tems (Dragan et al., 1995; Kim et al., 1996), and tamoxifen and its congenors induce aneuploidy in rodent tissues in vivo (Sargent et al., 1996). Thus, at least in some experimental systems, tamoxifen exhibits the characteristics of a complete carcinogen.
Figure 11.11 Breast cancer in women younger than 45 years of age: relative risk by total years of oral contraceptive use before first-term pregnancy in nine studies. OC, oral contraceptive; FTP, first-term preg- nancy. (Adapted from Schlesselman, 1990, with permission of the author and publisher.)
Androgenic steroids in the form of synthetic congeners of testosterone, such as oxymetholone, have been used in the long-term therapy of aplastic anemia. A number of case reports of hepatocellular carcinomas have arisen in individuals treated for this and other condi- tions by this form of steroid (cf. Hoover and Fraumeni, 1981).
A number of agents known to damage DNA have been used in the therapy and diagnosis of various diseases, especially neoplasia. Alkylating agents are used primarily for the treatment of malignant neoplasms, and a number have been causally associated with the induction of a variety of leukemias and solid neoplasms, especially in children receiving intensive therapy for acute leukemia (Chapter 20). The induction of acute nonlymphocytic leukemia has been re- ported in adults receiving such chemotherapy, especially for Hodgkin disease (Swerdlow et al.,1992; Karp and Smith, 1997). Table 11.14 lists a number of chemotherapeutic agents used in the treatment of cancer that have been evaluated in the IARC monograph program. As noted from the table, almost all of these agents are alkylating agents, which are thus capable of reacting with DNA in their ultimate forms (Chapter 3).
The risk of development of a secondary leukemia does not increase linearly with time but rather exhibits a sharp peak or “critical window” of incidence, as noted in Figure 11.12. Al-
Table 11.14 Selected Chemotherapy Agents That Have Been Evaluated in the IARC Monographs Program
Figure 11.12 Risk function for the development of leukemia in patients with Hodgkin disease regard- less of the type of treatment employed. The natural log transformation of the risk equation and the two- sided p value for the nonlinear term are noted in the figure. (Adapted from Blayney et al., 1987, with per- mission of the authors and publisher.)
though ionizing radiation is also utilized in the therapy of neoplasia, especially in association with chemotherapy at times, the carcinogenic effects of the chemotherapeutic agents can be dis- tinguished from those of ionizing radiation (cf. Swerdlow et al., 1992; Reimer, 1982). More re- cently, newer therapeutic agents are being employed, and, although studies are yet incomplete, evidence strongly suggests that the effective platinum drugs as well as the topoisomerase inhibi- tors are carcinogenic (Greene, 1992; Karp and Smith, 1997). The topoisomerase inhibitors in- duce specific chromosomal abnormalities that are reflected in the ultimate neoplasms, usually leukemias, suggesting that such agents are primarily effecting the stage of progression.
There is no question of the association of lung cancer with chronic exposure to arsenic in industrial situations, but the exposure to arsenic compounds in the treatment of various diseases was widespread during the earlier parts of this century. Organic arsenicals were used to treat syphilis, and there is some evidence for an association with skin cancer in individuals receiving such treatment for prolonged periods. However, a much firmer association of skin as well as liver cancer with the chronic administration of Fowler’s solution has now been well documented (Pershagen, 1981). This medicament, in the form of a solution of 1% potassium arsenite in aque- ous alcohol, was used for the treatment of dermatitis, arthritis, and other conditions, including chronic leukemia. Since the material was often administered for years, some patients received many grams of arsenic during their period of treatment. An interesting study by Fierz (1966) reported a dose-response relationship between the development of skin cancer and the ingestion of Fowler’s solution (Figure 11.13).
Chlornaphazine, which had been used to treat preleukemic conditions in humans, was shown to be carcinogenic for the bladder and today is no longer utilized in such therapy (IARC,1982). Methoxypsoralen, which interacts with DNA, has been used in combination with ultravi- olet light in the treatment of a skin condition known as psoriasis. Clear evidence of the induction of squamous cell carcinoma of the skin by this regimen has been reported (Stern and Laird,
1994). In addition, the administration of methotrexate, an analog of the vitamin folic acid, also increases the risk of skin cancer in patients with psoriasis treated by this antimetabolite. More recently, these same studies have indicated that the risk of malignant melanoma also increases in
Figure 11.13 Percentage frequency of skin carcinoma as a function of the total dose of Fowler’s solu- tion. (After Fierz, 1966.)
patients who have been treated with ultraviolet light and methoxypsoralen for extended periods, usually involving more than 250 separate treatments (Stern et al., 1997). There has also been some epidemiological evidence that extensive exposure to fluorescent light may be a potential risk factor for melanoma (Walter et al., 1992), and there is an even greater risk from the use of sunlamps for tanning in a Swedish population, with a relative risk factor greater than 7 when such devices are used more than 10 times a year by young people (Westerdahl et al., 1994). These instances were in the absence of treatment with methoxypsoralen, indicating, as reiterated in the next chapter, the potential danger of ultraviolet and near-ultraviolet radiation in the devel- opment of skin cancer in the human.
Although not strictly a chemical cause of human cancer, thorium dioxide in a colloidal form known as Thorotrast was once used for diagnostic imaging of the liver and other tissues in vivo for the diagnosis of both primary and metastatic neoplasia in the human. This radioactive compound is sequestered by tissues of the body, especially in the liver, and not released for long periods. The continuous radiation by alpha particles, as well as ionizing photons emitted from the spontaneous disintegration of thorium within the tissue, leads to neoplasia, especially of vas- cular elements in the liver, resulting in angiosarcoma (van Kaick et al., 1978).
A group of drugs has been utilized in the chemical immunosuppression of patients in prep- aration for transplantation of organs and tissues from one individual to another. The principal drug that has been associated with neoplasms in this group of patients is azathioprine. Although this drug and others used during clinical immunosuppression interact with DNA, it is not clear whether such compounds are actually carcinogenic in themselves or act by suppressing the natu- ral resistance of the host to malignant cells already present. The predominant neoplasms appear- ing in such immunosuppressed patients are those derived from the skin and the immune system. In a study of more than 6000 individuals having had kidney transplant surgery in the United States, Hoover and Fraumeni (1973) reported dramatic increases in the risk for these two types of neoplasms; the most striking was reticulum cell sarcoma, the occurrence of which was 350 times greater than in the general population. A more recent series from England (London et al.,1995) showed a somewhat different distribution of neoplasms in 70 of 918 renal transplant pa- tients; 53% of these tumors were cutaneous in origin. The risk of developing these neoplasms of the skin, mostly squamous cell carcinomas, rose almost exponentially during a 20-year follow- up. It is reasonable to suggest that a major mechanism for the induction of neoplasia in immuno- suppressed patients is the loss of host resistance to neoplastic cells that are already present but are normally prevented from expressing their neoplastic potential by immune mechanisms of the host. A marked increase in the incidence of Kaposi sarcoma has been seen in immunocompro- mised patients in this and other countries, especially male homosexuals with the acquired immu- nodeficiency syndrome (AIDS), a disease associated with marked immune dysfunction and loss of resistance to infectious agents (Kahn and Walker, 1998) (Chapter 12). These increases in neo- plasms seen in immunosuppressed individuals may in part reflect the inability of the host to re- sist the development of neoplasia by immunological mechanisms, as discussed further in Chapter 19.
The association of leukemia with administration of the antibiotic chloramphenicol may be related to its depressive action on the bone marrow (IARC, 1982). Similarly, in 1987, the IARC classified the analgesic phenacetin as a cause of cancers of the renal pelvis, ureter, and bladder. Today this agent is not available as an over-the-counter analgesic in most western countries. There is little or no evidence that salicylates or acetaminophen exhibit any carcinogenic effects in the human. The association of lymphomas with the chronic administration of phenytoin for the control of epilepsy is very limited, with significant conflicting evidence (Olsen et al., 1989; Scoville and White, 1981). This drug induces lymphoid reactions that are at times difficult to distinguish from neoplasia, but there is limited evidence for its causal association with lym- phoma, although the risk is probably much less than the benefit to the epileptic patient. An inter- esting but as yet unknown effect of administration of a known carcinogenic agent to humans for medicinal purposes was the use of ethylcarbamate (urethane) as a cosolvent in aqueous solutions of water-insoluble analgesics in Japan from 1950 through 1975 (Miller, 1991). No epidemiolog- ical follow-up was made of the use of some hundred million 2-mL ampules of 7% to 15% solu- tions of ethylcarbamate injected into patients between 1950 to 1975. Whether this and other medical treatments with potential carcinogens will alter cancer risk and incidence remains for future study.
Thus, it is clear that some forms of medical therapy and diagnosis are not without signifi- cant carcinogenic risk under certain circumstances. Chapter 13 presents a further discussion of the relation of benefit to risk in our society in the employment of known or suspected carcino- genic agents. In the final analysis, society and/or the individual must determine when one factor outweighs the other.