ETIOLOGY OF MALIGNANT DISEASE
WILLIAM A. BLATTNER
Over the last 20 years, public awareness of environmental factors, personal habits and lifestyle, and genetic predisposition associated with heightened cancer risk has translated into increased activism reflected in legislation, litigation, and personal commitment to a prevention-oriented lifestyle. Following the lead of AIDS activists, the demand for increased research funding, legislation to improve worker conditions and the environment at large, the “tobacco settlement” for health-related outcomes of tobacco use, and increased attention to diet and exercise are manifestations of this activism. The practicing clinician and subspecialist are called on to be the arbiter of an informed and balanced perspective on these issues.
Recent advances in molecular biology have improved our understanding of the genetic basis of cancer as a complex series of diseases involving multiple steps and multiple pathways involving genes at the center of cell signaling, DNA repair, and other vital check points of cellular function. Beginning with an initiation event that renders a tissue premalignant, followed by a number of promotional steps that increase the potential for an initiated cell to become malignant, cancer is a multistage process. Opportunities for early detection have increased with the emergence of screening approaches such as prostate-specific antigen (PSA), improved imaging techniques, and less invasive sampling such as fine-needle biopsy. These and improved therapies have started to make an impact on the morbidity and mortality of some tumors. With 70% of all deaths caused by cancer and cardiovascular disease in persons over age 65, declines in deaths due to cardiovascular disease and stroke are occurring at a rate faster than those for cancer. With the specter that cancer will become the leading cause of death in this age group in the United States and other developed countries in the next 5 to 10 years, identification of preventable exposures and strategies to reduce or reverse the risk for cancer remains a high priority.
In their 1981 monograph, Doll and Peto estimated that at least 30% to 40% of all cancer deaths could be avoided by applying the knowledge we have of the known causes of cancer. Effecting behavioral changes to reduce cancer risk is complex (Chapter 22), although major national campaigns in the areas of diet, physical exercise, and antismoking have achieved some success in changing behavior among targeted populations.
In the United States, tobacco use accounts for about 40% of all cancers among men and 20% of all cancers among women. Lung cancer is the major tobacco-associated cancer site, with 90% of cases among men and 79% of cases among women attributed to tobacco exposure. The epidemic of lung cancer in the United States emerged among men in the 1930s and among women in the mid-1960s—approximately 20 years after the widespread introduction of cigarette smoking in each of these groups. Since then, an overwhelming body of evidence has amassed that shows that cigarette smoking causes a variety of malignancies of the respiratory, gastrointestinal, and genitourinary systems. For lung cancer, the risk from heavy cigarette smoking (more than two packs per day) is 20 times higher than among nonsmokers. Filter-tipped cigarettes, which decrease the tar and nicotine levels, reduce the risk for smokers, but the rates are still much higher than among nonsmokers.
Nonsmokers exposed to environmental tobacco smoke, passive smoking, experience excess risk and present with patterns of tumors similar to that of smokers. Heavily exposed passive smokers have risk estimates similar to that of light smokers. Former smokers experience a reduction in risk compared with active smokers; this is detectable within a few years of cessation.
Other malignancies associated with cigarette smoking are by system: respiratory (lip, oral cavity, pharynx, larynx, trachea, bronchus, and lung); gastrointestinal (esophagus and pancreas); genitourinary (bladder, kidney, and ureter); reproductive (uterus, cervix); and hematologic (myelogenous leukemia). The risk of cancers of the lip, mouth, tongue, pharynx, larynx, and esophagus is further amplified by heavy alcohol consumption.
Smokeless tobacco usage, tobacco chewing, and particularly snuff dipping is of increasing concern, because the practice has become popular among teenagers and young adults. Rates of mouth and throat cancer are increased up to 50 times in long-time snuff users. In general, a dose-response relationship governs mouth and throat cancer; risks increase with the amount and duration of tobacco use.
Alcohol consumption synergistically enhances the risk for several tobacco-related neoplasms. Combined exposure to alcohol and smoking account for about 75% of all oral and pharyngeal cancers. Alcohol alone is estimated to contribute to about 3% of cancers and has been associated with colon cancer and colorectal adenomatous polyps and with esophageal, pancreatic, nasopharyngeal, prostate, and breast cancers. Ethanol is not carcinogenic in laboratory animal studies but may enhance carcinogenicity by making carcinogens more soluble or by facilitating tissue penetration. Alcohol is associated with heightened risk for liver cancer, and tissue injury resulting from cirrhosis may contribute to the carcinogenic process. Nutritional deficiencies among black men who are heavy drinkers have been linked to high rates of cancer of the esophagus. Several analyses suggest that even moderate use of alcohol enhances female breast cancer by a factor of 50%. Although the basis for this determination is unknown, effects on hormonal metabolism have been postulated.
DIET AND NUTRITION
Establishing clear links between diet and cancer is complicated by methodologic issues, such as differential recall bias and the long latency between putative dietary exposures and subsequent cancer risk. Observational studies, such as those among migrants from areas with low colon cancer risk who adopt a Western diet, thus heightening their risk for this tumor in their lifetime, demonstrate the importance of diet in cancer risk. Interventional studies, including one that used vitamin A supplementation to effect reversal of preneoplastic lesions of the esophagus in a high-risk Chinese population, mirror results in experimental animal studies in which supplemental vitamin administration reversed preneoplastic lesions. For cancers with a suspected dietary component, some have estimated that 35% of cancer has a dietary factor involved in the cause, but with a wide range for this attribution. Current public health guidelines suggest that a prudent diet, which emphasizes reduction in total animal fat consumption, increase in intake of fruits and vegetables, good food preservation and healthy preparation, moderation in alcohol consumption, and avoidance of obesity, may reduce cancer rates by as much as one-third while reducing risk of cardiovascular disease as well.
The mechanism or mechanisms by which diet may influence cancer risk—either by increasing or decreasing it, are complex, and it is difficult to distinguish which nutrients are the active agents and how much daily intake is required to have a desired effect. Obesity is strongly linked to risk of endometrial cancer and cancer of the biliary system, colon cancer in men, and possibly renal cell cancer. Higher mortality for breast cancer among obese postmenopausal women may result from delayed detection. However, high fat intake is associated with increased risk for breast cancer in correlational but less clearly in analytic studies. Prostate and colon cancers are more strongly associated with dietary fat and meat consumption, and some studies also suggest a role in ovarian cancer as well.
Sir Dennis Burkitt popularized consumption of high dietary grain fiber as a means of reducing colon cancer risk, but a recent study, while showing benefit of increased fiber intake for cardiovascular disease, did not make a favorable impact on colon cancer. However, high vegetable and fruit intake is beneficial, but the effect may result from vitamins or other nutrients (e.g., vitamin C, indoles, and other nutrients) in reducing cancer risk. Vitamin A, carotenoids, vitamin C, vitamin E, and selenium have been implicated in a variety of studies as having a beneficial effect in reducing cancer risk. These dietary adjuncts have been used in interventional studies in varying combinations or alone with mixed results, including the finding in one Finnish study in which men at high risk of lung cancer experienced increased rates of lung cancer when given vitamin A supplements. In this same trial, those receiving vitamin E did not experience an increased risk of lung cancer, but a modest reduction in prostate cancer was observed.
In vitro studies of carcinogens and mutagens detected as a result of food preparation (e.g., high-temperature cooking) raised questions about the role of these factors in human cancers. But epidemiologic studies in human populations have not linked such factors, except for salt-preserved foods that may heighten risk for stomach cancer and nasopharyngeal cancer in China. Several natural products are themselves carcinogenic, such as aflatoxin, a carcinogenic metabolic product of the fungus Aspergillus flavus, which is associated with high rates of liver cancer.
Compelling data from many sources document that diet may modify cancer risk by years. As a fuller understanding of these complex interactions develops and accumulated supportive evidence increases, acceleration of educational campaigns on the health benefits derived from improved nutritional practices may lead to substantial reduction in cancer rates.
The major classes of medically prescribed drugs linked to heightened risk for cancer are hormones, anticancer drugs, and immunosuppressive agents. Despite these associations, drugs are believed to account for less than 2% of all cancers.
In the late 1960s, an epidemic of vaginal and cervical adenocarcinoma in young women was linked to prior exposure in utero to diethylstilbestrol, which was used as an adjunct to preventing miscarriage. The use of synthetic conjugated estrogens alone for menopausal symptoms has been linked to the development of endometrial cancer, but low-dose estrogen replacement in combination with progesterone-containing hormonal replacement regimens is not associated with increased endometrial cancer risk. The relation of hormonal replacement therapy and risk for breast cancer is complex, with a number of studies conducted in the 1990s demonstrating modest increases in risk but with several studies suggesting that the associated subtype of breast cancer has a good prognosis. So the impact of such therapy on mortality is not known. Some have argued that the benefits of hormonal replacement outweigh the risks because of reduction in heart disease and osteoporosis. Decisions about hormonal replacement therapy depend on the individual circumstance of the patient and her medical needs as well as results of further studies that can help to refine our understanding of this complex paradigm.
There are conflicting data on oral contraceptives as a risk factor for breast cancer; early, prolonged use in persons with a predisposition is implicated. Other studies of oral contraceptives suggest that they may decrease the risk of ovarian and endometrial cancer, although an association with invasive cervical cancer was suggested in one study.
The concept of immunosurveillance in cause of cancer, first proposed in the early 1960s, suggested that cancer emerges because of a loss of immunologic recognition of tumor-associated antigens. Studies of immunodeficiency in various settings have documented a limited repertoire of tumor types. One of the first settings evaluated involved the study of tumors among patients receiving immunosuppressive therapies to suppress rejection of organ transplants. In this setting, non-Hodgkin’s lymphomas, especially of the brain, are the most prominent manifestation. The emergence of such tumors within months of transplantation contrasts with the longer duration for cancer induction associated with environmental carcinogens. An infectious agent, Epstein–Barr virus (EBV) is linked to such transplant lymphomas.
Other cancers associated with immunosuppressive therapy include Kaposi’s sarcoma and cervical, vulvar, and anal (all human papillomavirus–associated) cancers. Squamous cell carcinoma of the skin and malignant melanoma are also increased in this setting. This pattern of tumors has also been observed in acquired and congenital immunodeficiency, with some additional tumor types linked to specific congenital immunodeficiency states such as ovarian dysgerminomas, and stomach and liver cancers in ataxia telangiectasia, probably associated with this disorder’s chromosomal changes. The stomach cancer observed in common variable immunodeficiency is probably a result of the common occurrence of achlorhydria in that disorder.
With the exception of hormones and several medicinal agents, most drugs as they are ordinarily used clinically pose no risk for cancer. Arsenicals are no longer used in clinical practice but are associated with some risk for skin cancer. Diphenylhydantoin is linked to a slightly increased risk for non-Hodgkin’s lymphomas. Phenacetin used in high dosages is linked to renal cancer, whereas nonsteroidal anti-inflammatory drugs, including aspirin, have been shown to reduce the risk of colon cancer. Some anticancer drugs, particularly alkylating agents, which have radiation-like effects, are associated with increased cancer risk, particularly leukemia. However, their use is justified when treating otherwise incurable cancers. Studies of patients treated with alkylating agents for Hodgkin’s and non-Hodgkin’s lymphoma, multiple myeloma, and ovarian, gastric, and colorectal cancers have shown a 16- to several hundred-fold increase in risk for acute nonlymphocytic leukemia emerging after 3 to 5 years and peaking after 10 to 15 years.
Ionizing radiation produces its carcinogenic effects by direct damage to the genetic material of the cell. Radiogenic cancers are most prominent in the breast, brain, thyroid, and bone marrow, with excesses of some other tumors associated with particularly heavy local exposure to a particular site, such as with osteosarcomas and bone-seeking radionuclides.
Most data on the carcinogenic potential of ionizing radiation are derived from moderate to high exposure levels; extrapolations from these data suggest the importance of the cumulative effect of exposure. It is unlikely that a threshold dose exists below which there is no carcinogenic effect.
Radiogenic leukemia differs from most radiation-induced cancers in that the latent period is relatively short. Cases occur within a few years of exposure, peak at 6 to 8 years, and are followed by a decline to normal rates within 25 years. Radiogenic carcinomas have much longer latent periods. In the prospective follow-up of the children exposed to radiation after the atomic bomb blasts in Japan, prepubertal girls exposed to moderate doses were at high risk for breast cancers 20 to 30 years later. Even modest doses of prenatal radiation are associated with heightened risk for leukemia and other childhood cancers. Cancers have also developed in sites where radionuclides are concentrated. Some examples include osteosarcoma (radium 224), leukemia (phosphorus 32), and liver angiosarcoma (Thorotrast).
Although ionizing radiation appears to account for no more than 3% of all cancers, there is considerable public concern about this risk factor. Increased public awareness of radon in ground water and its potential for household contamination and the publicity surrounding the general population exposure from the nuclear accident in Chernobyl, intensify this concern.
Solar radiation causes up to 90% of nonmelanoma skin cancer, and its effects are linked to skin melanoma as well. The link between sunlight exposure and squamous and basal cell carcinomas was determined from the high rates among persons with outdoor occupations (e.g., sailors, farmers), among persons residing in southern latitudes, and among fair-skinned people with lower levels of protective melanin pigment.
Skin cancers tend to occur most prominently in sun-exposed areas. For nonmelanoma skin cancer, risk is related to annual cumulative lifetime ultraviolet-B exposure; for melanoma skin cancer, a history of repeated sunburn, especially in youth, is associated with heightened risk for subsequent skin melanoma years later. Behavioral modification, such as avoidance of excessive sunlight exposure and the use of readily available and highly protective sunscreens, could have a major preventive impact on cancers of the skin.
Exposures to potential carcinogens in the workplace have led to the recognition of several compounds as human carcinogens, such as asbestos, which causes mesothelioma, and vinyl chloride, which causes liver angiosarcoma. Workers exposed to aromatic amines in dye, rubber, and coal gas manufacture and some chemical workers are at increased risk for bladder cancer.
Respiratory carcinogens include bis-(chloromethyl)-ether (causes oat cell carcinoma of the lung), chromium manufacture (lung cancer), mustard gas exposure (lung, larynx, and nasal sinuses cancer), nickel dust exposure (tumors of lung and nasal sinuses), isopropyl alcohol production (tumors of nasal sinuses), polycyclic hydrocarbons (lung cancer), and wood dust in furniture manufacture (tumors of nasal sinuses). Benzene exposure in leather, petroleum, and other industries is linked to nonlymphocytic leukemia. Herbicide exposures among foresters and farmers are linked to lymphoproliferative and soft-tissue neoplasms. Cadmium exposure is associated with a heightened risk for prostate cancer, and formaldehyde exposure has been linked to nasopharyngeal cancer. Although occupational exposures account for as much as 5% of all cancer deaths, cancer prevention has been significantly advanced through the identification and elimination of hazardous exposures in the workplace.
The level of cancer risk attributable to air and water polluted with known carcinogens remains controversial. Results from studies of air pollution by specific manufacturing processes, such as smelter emissions of arsenic, are associated with localized increased risk for lung cancer. Although rates of lung cancer are higher in urban than in rural areas, studies that control for smoking and occupational risk factors and those based on estimates of higher exposure rates in the workplace do not indicate a major risk for air pollution. Between 1% and 2% of cancers are estimated to be due to past exposures, whereas based on existing data, less than 1% of future cancers will result from current air pollution levels.
Water pollution was identified as a potential source of cancer risk, recognizing that the process of chlorination produces trihalomethanes, which are carcinogenic and mutagenic. Results are inconclusive, but excessive levels of these substances have been shown to correlate with bladder, colon, and rectum cancer rates. Analyses suggest a significant correlation between volume of water ingested and risk for cancer. Ground water contamination from local toxic waste disposal and dumping has been implicated in some cancer clusters, but aside from drinking water containing unusually high levels of carcinogens, little substantive evidence suggests that drinking water contributes substantially to cancer risk.
The contribution of infectious agents to the pathogenesis of cancer varies considerably by geographic area and population, with approximately 10% to 15% of all cancers associated with infectious agents. In developed countries, the contribution made by infectious causes to the overall cancer burden is relatively low; in the populations of some developing countries of Africa and Asia, more than 50% of the cancers are linked to a viral agent.
The major infectious carcinogens worldwide are the hepatitis B and C viruses, which are strongly linked to hepatocellular carcinoma, the leading cause of cancer mortality in many areas where early life exposure confers a substantial risk. EBV, a herpesvirus, has been linked to Hodgkin’s and some non-Hodgkin’s lymphomas, especially Burkitt’s lymphoma and nasopharyngeal carcinoma (NPC). In southern China, a specific immunogenetic marker identifies persons at high risk for NPC. A variety of cofactors associated with altered immunity, particularly bouts of malaria, are thought to be necessary to trigger the endemic form of Burkitt’s lymphoma in Africa. EBV has also been identified in approximately 7% of gastric carcinomas but its etiologic role is not established.
A newly discovered gamma herpesvirus—HHV-8, also known as Kaposi’s sarcoma associated virus (KSHV)—is linked to Kaposi’s sarcoma in AIDS and in an AIDS-associated form of extranodal (i.e., body cavity) non-Hodgkin’s lymphoma as well as the premalignant lymphoproliferative disorder, Castleman’s disease.
Cervical and vulvar cancer in women and penile cancer and anal carcinoma in men are linked to some subtypes of human papillomavirus (HPV). Prospective studies demonstrate that high levels of HPV in prediagnostic Papanicolaou (Pap) smears are elevated in some cases decades before the emergence of cervical cancer. This is consistent with the hypothesis that inability to down-regulate papillomavirus expression predisposes to subsequent cervical cancer risk.
The first human RNA retrovirus, human T-cell lymphotrophic virus type I (HTLV-I), is associated with a distinctive clinical pathologic entity, adult T-cell leukemia-lymphoma (ATL). Rarely, in persons who have had early-life exposure, a clonally integrated neoplasm occurs. The disease and virus cluster in southern Japan, the Caribbean basin, and surrounding countries; in these geographic areas, ATL is responsible for more than 50% of all lymphomas, which account for 1% to 2% of all malignancies in these countries.
Because of its immunosuppressive effects, HIV-1, the causal agent of AIDS, is associated with an increased risk for Kaposi’s sarcoma, non-Hodgkin’s lymphoma, Hodgkin’s lymphoma, and anal and possibly cervical carcinomas. With the advent of highly active antiretroviral therapy, the incidence of Kaposi’s sarcoma has dramatically declined and regressions of existing lesions noted. The impact of such therapy on non-Hodgkin lymphoma is less certain.
Certain nonviral parasites are also carcinogenic. Schistosomiasis is associated with squamous cell carcinoma of the bladder in the Middle East and North Africa, and liver flukes that cause clonorchiasis and opisthorchiasis are associated with cholangiocarcinoma in Asia.
Some bacterial agents, such as Helicobacter pylori, have been linked to gastric carcinoma and mucosa-associated lymphoid tissue lymphoma (MALT).
Although the role for environmental factors in producing cancer is beyond dispute, genetic factors are also consequential. Studies of human populations confirm a wide range of variability in inherent sensitivities to carcinogenesis. The role of individual variation is advanced by the disparity in cancer rates among distinct racial and ethnic populations in which environmental influences cannot account for the differences. For example, chronic lymphocytic leukemia is consistently absent in Asian populations, regardless of geographic locale; testicular cancer and Ewing’s sarcoma occur rarely in blacks, whether they reside in Africa or in the United States.
Molecular epidemiologic techniques are being applied to high-risk groups (a) to examine individual genetic variations that may be identified with increased risk, including variations in the pathways for metabolizing endogenous and exogenous, carcinogenic compounds; (b) to evaluate chromosomal instability; and (c) to investigate oncogenes and suppressor genes. Specific genetic changes have been identified as critical molecular events in the initiation and development of many cancers. Some of these structural changes include activation of oncogenes, inactivation of tumor suppressor genes such as P53 and the retinoblastoma gene (RB1), and chromosome deletions (Table 19.1).
TABLE 19.1. MOLECULAR BASIS OF SELECTED CANCERS AND SYNDROMES
Over 200 single-gene disorders have been recognized, which confer a cancer-prone genetic predisposition. Studies of such high-risk patients with these predispositions have yielded important insights concerning the fundamental biology of cancer. Hereditary neoplasms are best exemplified by autosomal dominant gene disorders, which result in the development of specific neoplasms or constellations of tumors. Approximately 40% of all retinoblastomas occurring in childhood are of the hereditary type. A deletion involving the long arm of chromosome 13 has been recognized in some cases, and molecular analyses with genetic probes have identified a specific gene region called a suppressor gene, which acts by modulating specific cell-cycling regulators. The absence of this down-regulation results in increased and unregulated cell proliferation. Thus, mutations in such genes—and in the case of retinoblastoma a specific gene, RB1, a DNA-binding nuclear protein—confer a high risk for tumor formation. For such suppressor genes to lose activity, both alleles of the gene must be mutated. In the case of hereditary retinoblastoma, in which patients are also prone to osteosarcoma of the leg and radiogenic sarcoma of the orbit, susceptible patients are born with an inherited constitutional rearrangement or deletion of chromosome 13q14, where the RB1 gene resides. A single mutation of the remaining intact RB1 gene allele results in tumor formation.
Studies of sporadic oat cell carcinoma of the lung and renal carcinoma also suggest a mechanism similar to retinoblastoma that involves a gene on the long arm of chromosome 3. Other cancer genes are the targets of gene-mapping studies applied to a variety of cancer-prone disorders. Table 19.1 lists some of the cancers for which studies of genetic or familial syndromes have contributed to understanding their molecular basis. The practice of oncology will be significantly influenced by these discoveries as these markers are applied to the diagnosis and staging of a wide range of common cancers.
A variety of preneoplastic states have been recognized. Hamartomatous syndromes are typified by autosomal dominant disorders in which faulty embryonic development results in localized abnormal growth in mixed component tissues and heightened risk for various cancers. The genodermatoses are autosomal recessive disorders linked primarily to skin cancers, particularly of sun-exposed areas. Defects in the repair of ultraviolet-induced DNA damage in xeroderma pigmentosum have provided insights into the types of metabolic pathways by which the body repairs solar radiation-induced damage.
The dysplastic nevus syndrome predisposes to malignant melanoma, and because it occurs as a heritable and sporadic condition, it is an important precancerous condition that practitioners are likely to see, diagnose, and cure.
Congenital immune deficiency states predispose to a variety of cancers, especially of the lymphoreticular system. Ataxia telangiectasia, an immune deficiency syndrome, also shares a defect in chromosome fragility. Tissue from these patients is usually susceptible to g-radiation exposure in vitro and in vivo. Patients with this autosomal recessive disorder are prone to lymphoma, lymphocytic leukemia, stomach cancer, and other cancers; heterozygous relatives are said to be at heightened risk for leukemia, lymphoma, and carcinoma of the biliary tract and a variety of other cancers.
The finding of cancer in close relatives is not unusual, and the risk for a particular cancer has been consistently reported to be about two- to threefold if a close family member has that tumor. Familial cancer family syndrome patients are often distinguished by a tendency for multiple primary cancers in the same person and often by a younger than usual age of onset. Site-specific familial cancer aggregations are the most common, with familial breast and colon clusters observed most frequently. In other instances, multiple types of cancer occur in the same family. Examples include the multiple adenocarcinoma syndrome (colon, endometrial and breast carcinoma); Turcot’s syndrome (brain and colon cancer); and the Li–Fraumeni syndrome (bony and soft tissue sarcomas, breast, brain, lung, larynx and adrenocortical neoplasms, and leukemia).
Cancer is a complex disease involving multistep molecular and cellular processes. Because no single genetic factor is sufficient to predict risk, the ultimate goal of epidemiology is to understand the environmental factors, lifestyles, and individual risk profiles that can facilitate population-based and individually targeted prevention approaches.
Some cancers are already declining as a result of efforts to eliminate exposures such as cigarette smoking. Nonetheless, more research is needed to identify the risk factors for common cancers, particularly those that are increasing, such as breast cancers in young women and non-Hodgkin’s lymphomas.
Collaborative Group on Hormonal Factors in Breast Cancer. Breast cancer and hormonal replacement therapy: collaborative reanalysis of data from 51 epidemiologic studies of 52,705 women with breast cancer. Lancet 1997;350:1484.
Doll R, Peto R. The cause of cancer. New York, Oxford University Press, 1981.
Evans A, Kaslow R, eds. Viral infections of man. New York: Plenum Press, 1995.
Gapstur S, Morrow M, Sellers T. Hormone replacement therapy and risk of breast cancer with a favorable histology. JAMA 1999;281:2091.
Perera F. Environment and cancer: who are susceptible? Science 1997;278:1068.
Ron E. Ionizing radiation and cancer risk: evidence from epidemiology. Radiat Res 1998;150(5 Suppl):S30.
Schottenfeld D, Fraumeni JF Jr, eds. Cancer epidemiology and prevention. New York: Oxford University Press, 1996.
Vessey MP, Gray M. Cancer risks and prevention. Oxford: Oxford University Press, 1985.
Willett W, Colditz G, Mueller N. Strategies for minimizing cancer risk. Sci Am 1996;375:88.