The chemical induction of cancer in animals is thought to involve a multistage process with a long latency period. This process can be initiated by a variety of chemical structures that have at least one common thread in their mode of action: an interaction with DNA.⁷, ⁸, ⁹ Initiation results from irreversible genetic alterations, such as mutations or deletions in DNA.¹⁰ One of the most carefully studied systems of tumorogenesis is the induction of skin cancer in mice and rabbits by alkylating agents, polycyclic hydrocarbons, and ethyl carbamate. Repeated applications of these agents over long periods result in the development of benign or malignant tumors. Exposure to these compounds in limited doses, however, causes morphologic changes in the epithelium but does not result in tumors unless this stage of initiation is followed by the introduction of a promoter, such as a phorbol ester, an ingredient of croton oil. Promoters are not carcinogenic by themselves but lead to tumor production if applied after the initiating agent. Treatment with a specific promoter before exposure to the initiating agent does not result in tumor formation. This stage of promotion occurs over weeks and months and is reversible in its early stages. Promotion involves changes not in DNA structure but in the expression of the genome mediated through promoter-receptor interaction. The binding of promoter to receptor alters the expression of genes downstream. For example, estrogens and androgens may act as promoters by binding to the estrogen and androgen receptors, respectively, in liver or mammary tissue. The final stage of progression is irreversible and is characterized by karyotypic instability and malignant growth. Thus, carcinogenesis is a multistep process that may be arrested at intermediate stages, that requires a long latency period for induction, and that can be influenced by a promoting agent. For cancer patients, induction of second tumors may require not only an initiator (a DNA damaging agent) but also a promoter, a function
that may be fulfilled by a second chemotherapeutic agent, by radiotherapy, by a disease-related abnormality in metabolism or immune function, or by a somatic mutation or deletion that increases the risk of malignancy. Chemical carcinogens show a diversity of structures but share important metabolic features. Most are inert and require microsomal metabolic activation to positively charged (or electrophilic) intermediates that react with DNA bases. This characteristic of carcinogens is shared by certain antineoplastic agents such as cyclophosphamide, procarbazine, and mitomycin C and is essential in the antineoplastic action of these drugs. Other agents, such as L-phenylalanine mustard and nitrogen mustard, do not require metabolic activation to form alkylating species. Carcinogenicity has also been ascribed to ionizing irradiation, which produces free radicals, such as superoxide or hydroxyl radicals. A number of antitumor drugs have the same ability to promote formation of reactive oxygen intermediates, such agents include those that possess quinone functional groups (doxorubicin hydrochloride and plicamycin [mithramycin]) and those that bind electron-donating heavy metals (such as bleomycin sulfate and hydroxyurea).

Host factors, including enzymes such as glutathione S-transferases (GST), detoxify potentially mutagenic and toxic DNA-reactive electrophiles. Functional polymorphisms of GSTP1 (codon 105 Val allele) are associated with a higher risk of treatment (chemotherapy)-related acute myelogenous leukemia (AML). This risk is particularly relevant in patients receiving agents that are substrates for GSTP1.¹¹ Other polymorphisms of drug-metabolizing enzymes, including cytochrome P450 3A4, NAD(P)H:quinine oxidoreductase and myeloperoxidase, may be markers of susceptibility to genotoxicity.¹² Direct genotoxicity may not be the sole explanation for drug-induced carcinogenesis. Short and long interspersed elements (SINEs and LINEs) comprise one quarter of the human genome and are spread throughout the genome through retrotransposition. In retrotransposition, an element is transcribed into RNA and then converted back to DNA by reverse transcription. The copied DNA is then reinserted into a new location. Genotoxic agents and γ irradiation have been shown to induce SINE transcription and reverse transcriptase activity. This observation suggests that genotoxic exposure may lead to genomic mutation through both DNA damage and through potentially mutagenic mobile elements in the genome.¹³

The identification of oncogenes and suppressor genes has added another variable to the equation. Several possible mechanisms of actions have been proposed. The loss of one allele in a tumor suppressor gene such as p53 can potentially increase the risk of a drug-induced mutation in the other allele and development of the malignant phenotype. Oncogenes can be activated by a variety of mechanisms summarized in Table 42-1. Point mutations, chromosomal translocations, and gene amplification can alter expression of these genes. Just as altered oncogene expression and mutation occur with exposure to potential carcinogens, exposure to carcinogenic antitumor drugs likely alters oncogene expression and increases the risk of second malignancies.

Five test systems for carcinogen exposure are available. Mutagenesis assays such as the Ames test attempt to quantify the frequency with which a chemical induces mutational events based on the assumption that mutagenicity correlates with the likelihood of causing cancer in animals. In extensive testing of a wide variety of agents previously documented to be carcinogens and noncarcinogens, 90% of the known carcinogens gave positive results in the Ames assay and 87% of the noncarcinogens were inactive.¹⁴,¹⁵ Many of the antineoplastic agents in use today have been examined in the Ames system and some of the results are incorporated in Table 42-2.¹⁶, ¹⁷, ¹⁸, ¹⁹ At least seven classes of agents known to contain carcinogenic compounds are poorly detected in the Ames system, including azo compounds, carbonyl, hydrazine, chloroethylene, steroid, and antimetabolite structures. Cytogenetic studies attempt to correlate drug-induced chromosomal aberrations such as sister chromatid exchanges (SCEs) with carcinogenicity.²⁰ Assay of SCE, a type of chromosomal study that detects the exchange of small DNA fragments between sister chromatid pairs, has particular appeal because the effects of chemotherapeutic agents can be assessed in vivo by performing this test on peripheral lymphocytes from patients receiving antineoplastic therapy. Studies of lymphocytes from patients before and at intervals after chemotherapy have shown a marked increase in SCEs after the administration of lomustine (CCNU), dacarbazine, and mitomycin.²⁰, ²¹, ²²

Tests of oncogenesis in tissue culture are based on the hypothesis that agents that produce neoplastic transformation in culture are likely to be carcinogenic in the whole animal. This approach assumes that the drug concentration, duration of exposure, and metabolism of the suspected carcinogen are relevant to the in vivo situation, which may not be valid. Carcinogenicity studies in animals to predict the risk in humans are usually conducted in rodents over extended periods and at great expense. The primary drawbacks of this system are the known species, sex, and age dependencies of drug metabolism in rodents and the lack of pharmacologic information that would allow an extrapolation of results from rodents to humans.

The fifth approach, a measure of carcinogen exposure, uses detection of carcinogen-macromolecular adducts or somatic gene mutation in either target tissue or peripheral blood elements in animals or in man. The assumption here is that adduct formation or mutation will lead to cancer.

Although experimental evidence demonstrating the carcinogenic potential of many antineoplastic agents is abundant, the clinical evidence of this problem has been slower to appear. The fact that the rate of development of “secondary” cancers in patients with malignant lymphoma, pediatric cancers, ovarian cancer, and breast cancer is higher than that seen in an age-matched normal population has become clear. Many good reviews of this topic are now available in the medical literature.²³, ²⁴, ²⁵, ²⁶ Data reported more recently have come from hospital-based, national, and international tumor registries and from longer follow-up of chemotherapy and hormonal therapy studies.

One method that is used to identify treatment factors involved in the development of new cancers from a registry is the “nested” case-control study. Patients in the registry who develop a second cancer are compared with others who did not. These comparisons have provided a better estimate of the risks and identified factors that influence the development of second cancers. Clinical information about the total dose of drugs, concomitant therapy, and the duration of treatment is important in estimating risk. For some drugs such as the alkylating agents or etoposide, a threshold exists above which the risk of neoplasia rises sharply. Another issue in assessing the true risk of second cancers from cytotoxic agents is the existence of other factors that may also influence their development. An underlying increased incidence of second malignancy is found independent of therapy in patients with retinoblastoma, Wilms’ tumor, multiple myeloma, Hodgkin’s disease, and other tumors such as those associated with the hereditary nonpolyposis colorectal cancer syndrome. Other therapies used to treat the cancer, particularly radiation therapy, also impact the development of secondary cancers. An increase in solid tumors after therapy for Hodgkin’s disease and testicular cancer is most likely related to radiation rather than chemotherapy. In many reports, combination treatment regimens or regimens using irradiation and chemotherapy were used. Thus, the carcinogenic effects cannot necessarily be ascribed to one compound of the regimen with certainty, although the use of the nested case-control method may allow conclusions to be drawn regarding the carcinogenicity of different components of the regimen.

Interpretation of studies in this area must also take into consideration the statistical methods used to assess relative risk (RR).²⁷ The use of a person-years-of-risk analysis assumes that the yearly incidence of second malignancies is constant for the entire follow-up period and does not allow for the fact that a patient must live a certain
time through the latency period before the occurrence of a second malignancy. Such an analysis allows a reasonable estimate of the carcinogenic effects of a single therapy, but its use when comparing two treatments results in bias against the treatment that leads to a longer survival. Many studies compare the risk of cancer in the treated group with that of an age-matched cohort in the normal population to determine an RR. For a tumor that is uncommon in this age-matched population, a 5-fold to 10-fold increase in the RR sounds impressive but may only translate into a problem for fewer than 1% of patients who received therapy. On the other hand, small increases in RR for the more common solid tumors such as lung or breast cancer translate into a much greater problem in terms of absolute risk. One method that is useful in determining the overall impact of a secondary cancer in a population is to describe it in terms of the number of new cancers that occur per 10,000 patients treated.

Based on information currently available, one can categorize antineoplastic agents into high, moderate, low, and unknown risk groups on the basis of their oncogenic potential in humans (Table 42-3). This classification is based on reports of second malignancy in patients treated for both hematologic and solid tumors, with additional information coming from trials of cytotoxic agents in patients with immune diseases or after organ transplantation. Given that the latency period for the development of secondary cancers can range from 1 year (e.g., for etoposide-induced leukemias) to 20 years for solid tumors, the risk for many newer agents such as paclitaxel, docetaxel, irinotecan hydrochloride, gemcitabine hydrochloride, pemetrexed, and oxaliplatin cannot yet be properly determined. Furthermore, the impact of the new targeted agents such as imatinib, gefitinib, and erlotinib on formation of secondary malignancies is unclear. Neither gefitinib nor erlotinib has demonstrated genotoxic potential with in vitro or in vivo assays. Imatinib has been shown to induce benign and malignant tumors of preputial/clitoral gland, kidney, and urinary bladder in rats. This has not been demonstrated in humans.²⁸ A true assessment of agents primarily used in palliative therapy is also difficult because most patients may not survive long enough for problems such as second cancers to manifest.