Brain and Central Nervous System Cancer

Brain and Central Nervous System Cancer

Brain and Central Nervous System Cancer Melissa Bondy Randa El-Zein M.D. Anderson Cancer Center, Houston, Texas Margaret Wrensch University of Califo...

87KB Sizes 2 Downloads 38 Views

Brain and Central Nervous System Cancer Melissa Bondy Randa El-Zein M.D. Anderson Cancer Center, Houston, Texas

Margaret Wrensch University of California, San Francisco

I. II. III. IV. V. VI.

Germline p53 mutations have also been identified with this syndrome. glioma The general term for a tumor that arises from the supportive tissue of the brain (glial cells). Approximately 90% of brain tumors are gliomas. Examples of these include astrocytoma, oligodendroglioma, and glioblastoma multiforme. hereditary syndrome An inherited condition (or new mutation) that predisposes an individual to developing a brain tumor. Some of these syndromes include neurofibromatosis, tuberous sclerosis, and Turcot’s syndrome. mutagen sensitivity A measure of interindividual differences in DNA repair activity, measured by the rate at which chromatid breaks are repaired. Mutagen sensitivity is associated with an increased risk for cancer and may reflect the presence of an altered DNA repair pathway.

Introduction Congenital Conditions Familial Associations of Brain Tumors Genetic Susceptibility Etiology and Risk Factors Conclusions

GLOSSARY brain tumors Brain tumors are tumors that grow in the brain and can be either benign or malignant. A benign brain tumor consists of benign (harmless) cells and has distinct boundaries, but may occur in a vital area of the brain and be malignant. A malignant brain tumor is life-threatening, has uncontrolled growth, and poor prognosis. cancer family syndrome (Li–Fraumeni syndrome) A syndrome first described by Li and Fraumeni in 1969 showing a familial association of breast cancer, sarcoma, leukemia, and brain tumors. It has been shown that this association is vertically transmitted in a dominantly inherited pattern.

Encyclopedia of Cancer, Second Edition Volume 1

B rain cancer accounts for approximately 9% of all cancers and 90% of these are gliomas. The


Copyright 2002, Elsevier Science (USA). All rights reserved.



incidence of brain cancer has been rising, and much of this rise could be a result of more accurate and comprehensive diagnoses rather than changing or more extensive exposures. Although the etiology of brain tumors is unclear, environmental exposures, family history, and genetic susceptibility are possible explanations for this disease. This article reviews and discusses each of these.

I. INTRODUCTION Environmental concerns, media, and questionable research methodologies complicate progress toward understanding the etiology of brain cancer and obscure the truism that, at the molecular level, all cancer is genetic. We here first emphasize and discuss the achievement of consensus that the etiology of tumors requires random, inherited, or induced initiation in genes. We review putative risk studies, which point to a need for genetic epidemiology, with sensitive statistical methods as the best hope for an explanation of brain cancer etiology. Brain cancer, accounting for approximately 1.4% of all cancers and 2.3% of cancer deaths, is never truly benign, as slight impairments of brain or central nervous system (CNS) function, from cancer or its treatment, may have dramatic consequence. The incidence of brain cancer has been rising, and much of this rise could be a result of more accurate and comprehensive diagnoses rather than changing or more extensive exposures. Accurate and complete reporting and the construction of relevant databases required in epidemiology develop only slowly, as can be seen in the limited coverage or catchment area of the best U.S. collections of cancer cases. National monitoring of brain cancer is lacking, as are even universally agreed upon classifications of types and grades of invasiveness of brain tumors. Primary brain tumors are currently classified in a manner that reflects their histological appearance and location. The common typology includes the following. Gliomas arise from the glial tissues, accounting for 40% of CNS neoplasms and is a general category that includes astrocytomas, oligodendrogliomas, and ependymomas. According to the World Health Organization (WHO), there are four major grades of as-

trocytoma so denoted by their cellular star-like appearance and the most invasive of the tumors in children and adults. Astrocytomas are subdivided into grade I or pilocytic astrocytomas and are the most frequent brain tumors in children. These tumors rarely undergo neoplastic transformation. Grade II or low-grade astrocytomas account for 25% of all gliomas and are infiltrative in nature. Grade III or anaplastic malignant astrocytomas are highly malignant gliomas and have an increased tendency to progress to glioblastoma. Grade IV or glioblastoma multiforme is a highly malignant brain tumor and typically affects adults. This type of glioma has poor prognosis, largely because the tumor rapidly spreads to other regions of the brain. Oligodendrogliomas account for less than 10% of intracranial tumors. Oligodendrogliomas take their name and arise from the oligodendrocytes in the brain. Oligodendrogliomas are less aggressive than astrocytomas but are invasive and can traverse the cerebral spinal fluid (CSF). The ability of oligodendrogliomas to metastasize complicates their surgical removal, but because they are limited to the brain and CSF, some patients have a better prognosis and longer survival. Ependymomas are tumors arising from cells lining the brain ventricles or ependymal cells. Their growth may block the flow of CSF, causing notable swelling of the ventricle or hydrocephalus. Although ependymomas may move along the CSF, they characteristically do not infiltrate normal brain tissue and are sometimes amenable to surgical treatment, especially surgery of the spinal cord. Meningiomas arise from the sheaths surrounding the brain. The growth of meningiomas and the pressure they produce lead to the symptoms of brain tumors. Meningiomas are quite common, accounting for about 50% of primary central nervous system tumors. Because meningiomas are usually near the surface of the brain, they are often operable and are usually benign. Medulloblastomas include primitive neuroectodermal tumors that arise in the cerebellum. Medulloblastomas replicate quickly but can be treated with radiation because of their site specificity and early age of onset. They occur most commonly in children and frequently spread throughout the CSF.


Ganglioglioma are tumors containing both neurons and glial cells. They have a high rate of cure by surgery alone or surgery combined with radiation therapy. Schwannomas (neurilemomas) arise from Schwann cells, which surround cranial and other nerves. Schwannomas are usually benign tumors and often form near the cerebellum and in the cranial nerves responsible for hearing and balance. Chordomas are spinal tumors that preferentially arise at the extremities of the spinal column and usually do not invade brain tissues and other organs. They are amenable to treatment but stubbornly recur over a span of 10–20 years. Geneticists, molecular biologists, and epidemiologists are seeking methods for identifying and characterizing brain cancer genes for a clearer understanding of cancer etiology and to develop prevention strategies. Of special interest in carcinogenesis are protooncogenes, initiating carcinogenesis by activating cell division, and suppressor genes, inhibiting tumors. Epigenetic processes may also inhibit or stimulate tumor growth. Equally essential information under scrutiny is the mechanism of gene–environment interaction. Guiding characterization and gene– environment investigations are studies of known heritable syndromes associated with CNS tumors. These studies also indicate the potential of further collection and use of genetic data.

II. CONGENITAL CONDITIONS Original studies of CNS tumor-associated syndromes, as well as hereditary conditions associated with CNS tumors, parallel in method and implication the studies of other congenital anomalies. These, and followups, associate congenital medulloblastoma with gastrointestinal and genitourinary system anomalies, congenital ependymoma with multisystem anomalies, astrocytoma with arteriovenous malformation of the overlying meninges, and glioblastoma multiforme with adjacent arteriovenous angiomatous malformation and pulmonary arteriovenous fistula. CNS tumors commonly arise in individuals with Down’s syndrome, a disorder involving trisomy 21, and gliomatous tumors with syringomyelia, a disorder possibly of genetic origin. Mental retardation may also be


associated with familial brain cancer, as children with astrocytomas had a mentally retarded sibling three times more frequently than controls (p  0.04), whereas mentally retarded sibs, nieces, and nephews occurred in families of adult males 4.8 times more often than in families of controls. Many reports note cooccurrence of CNS tumors and hereditary syndromes. They are described briefly here. Tuberous sclerosis, or Bourneville’s disease, is an autosomal dominantly inherited progressive disorder occurring in 1 per 10,000–150,000 persons. It is characterized by hamartomas and hamartias of the skin, CNS, and kidneys and results in sebaceous adenomas of the skin, muscle and retinal tumors, epileptic seizures, mental retardation, and nodes or tubers of abnormal glial fibers and ganglion cells in the brain. Its association with CNS tumors is anecdotal, although one hospital study reported seven CNS tumors in 48 cases (15%) of tuberous sclerosis, whereas another found 22 cases of subependymal giant cell astrocytoma in 345 patients (6.4%) with tuberous sclerosis. Astrocytoma, ependymoma, and glioblastoma multiforme have been associated with tuberous sclerosis in up to 5% of cases. Contradictory studies, some reporting linkage to the long arm of chromosome 9q32-34 and others linkage to loci on 11q, indicate that tuberous sclerosis may be genetically heterogeneous. Neurofibromatosis (NF-1), or von Recklinghausen’s disease, occurring in 1 of 3000 live births and showing an autosomal-dominant pattern of heredity, is regarded as among the most common single gene disorders. Paternal origin of the mutation was found in one study in 12 of 14 families, but single gene etiology has yet to be firmly established, as the spontaneous mutation rate of large populations has been put at 50%. NF-1 characteristics are cutaneous pigmentation (cafe-au-lait spots) and multiple neurofibromas involving the skin and possibly deeper peripheral nerves and neural roots. NF-1 patients (94%) present with Lisch nodules or pigmented iris hamartomas and commonly experience optic nerve gliomas, astrocytomas, ependymomas, acoustic neuromas, neurilemmomas, meningiomas, and neurofibromas. Of NF-1 patients, 4–45% experience brain tumors. Neurofibromatosis type 2, or bilateral acoustic



neurofibromatosis, occurs with one-tenth the frequency of NF-1. NF-2 presents the clinical characteristics of multiple tumors, usually schwannomas of the cranial and spinal nerve roots. Multiple ependymomas, meningiomas developing from arachnoidal cells in the cranial cavity and spinal canal, and spinal cord or brain stem astrocytic gliomas occur in individuals with NF-2. These are often low-grade malignancy but with devastating neurological effects. NF-2 is caused by a deletion in the long arm chromosome 22 associated with meningiomas, gliomas, and spinal neurofibromas. Nevoid basal cell carcinoma syndrome, or Gorlin syndrome, an autosomal-dominant disorder, presents with multiple basal cell carcinomas having arisen early in life and jaw cysts, characteristic facies, skeletal anomalies, intracranial calcifications of the falx, and ovarian fibromas. The syndrome, but not its rate of coincidence, was associated with medulloblastoma. Loss of heterozygosity at the chromosomal location examined, particularly in hereditary tumors, implies that the gene, normally functioning as a tumor suppressor, is homozygously inactivated. Bare and colleagues found loss of heterozygosity on chromosome 1q22. Turcot’s syndrome, Gardner’s syndrome, and familial polyposis, characterized by adenomatous polyps, have been associated with medulloblastoma and glioblastoma. Because of their marked similarity, some authorities consider Turcot’s syndrome, Gardner’s syndrome, and classical adenomatous polyposis variations of a single genetic defect. Investigators associate chromosome 5q with these three syndromes. Sturge–Weber disease is an inherited neurocutaneous syndrome characterized by facial and leptomeningeal angiomas and, frequently, facial and optical port wine lesions. Computer-assisted tomography and magnetic resonance imaging of Sturge– Weber cases show cerebral lobar atrophy, brain calcification, choroid plexus enlargement, and venous abnormalities. Von Hippel–Lindau disease, a rare autosomaldominant multisystem disorder, involves cerebellar hemangioblastoma of the CNS and visceral organs, retinal angiomatosis, pancreatic cysts, and benign and malignant renal lesions. Several studies link von Hippel–Lindau disease to the short arm of chromosome 3 (3p13-14.3 and 3p25-26). These findings strongly suggest that a tumor suppressor gene is in-

volved in the disease, and discovery of the gene(s) should enable a reliable diagnostic test to screen family members.

III. FAMILIAL ASSOCIATIONS OF BRAIN TUMORS In addition to the association of hereditary syndromes with CNS tumors, investigation of brain cancer etiology focuses on families of CNS tumor patients aggregating CNS and other cancers. Tumors of patients and their relatives in these “cancer families” are histologically and biologically similar, and well documented, although the precise relationship between genetics and CNS neoplasms remains unknown. Methodologic constraints unfortunately limit the authority of many of these studies, also obscured by the confounding factor of common familial exposure to environmental agents potentially contributing to neoplasia induction, but they consistently report the presence of similar brain or other tumors in siblings and the cancer family syndrome. With regard to etiology, two possible explanations of family occurrence emerge: a genetic factor may in itself cause family clustering of CNS tumors or a hereditary vulnerability to exposures may produce the clusters.

A. CNS Tumors among Twins A problem in the area of family studies is the lack of, or yet undiscovered finding of, concordance of CNS tumors in twins. Incidence studies of concordance of twins with CNS tumors in twins have not shown that twins’ tumors are histologically identical. In a review of childhood cancer deaths in 145,708 twins and singletons born between 1940 and 1964, Norris and Jackson found 54 instances of solid tumors, 21 were brain tumors. Neither Harvald and Hauge nor Miller found evidence of concordance of CNS neoplasms in twins. To our knowledge, no recent reports discuss the etiologic significance of lack of concordance of CNS tumors in twins.

B. CNS Tumors among Siblings Surveys of childhood cancer registries reveal a high frequency of siblings concordant for brain tumors, sib-


lings with childhood brain tumors and leukemia, and siblings with brain tumors and other childhood cancers. Miller’s national survey of children’s death certificates from 1960 to 1967 found a significant excess of brain and bone cancers and leukemia among patients’ siblings. The lack of adequate numbers in some early surveys was overcome by Farwell and Flannery using the Connecticut Tumor Registry to compare cancer incidence in parents, siblings, and children of 643 patients with CNS tumors in childhood. They used controls selected from Connecticut birth certificate files matched to cases by sex, age, and birth place and found no case-control cancer risk differential but a significantly higher risk of brain cancer in case siblings. Similarly, Draper and colleagues used the Marie Curie/Oxford Survey of Childhood Cancers in England, Scotland, and Wales to review more than 20,000 cases of malignant neoplasms. They found 11 sib pairs with brain tumors and 21 sib pairs with dissimilar cancers. They, as the other sibling researchers, concluded that sibs of patients are at increased risk and, further, susceptibility to cancer might be genetic.

C. CNS Tumors and Cancer Family Syndrome Going beyond twin and sibling studies, family studies derived pedigrees of brain and other tumors consistent with a dominantly inherited disorder. In 1988, Li and co-workers described this cancer family syndrome in 24 kindred that had both childhood and adult onset cancers of diverse sites. Fourteen (9%) of the 151 cancers that occurred before age 45 were brain tumors. The Li–Fraumeni cancer syndrome has since been linked to a p53 mutation on chromosome 17p in some families. Additional evidence for a family syndrome comes from many epidemiologic comparisons of family medical histories of brain tumor cases to those of controls. These report a relative risk of 1 to 1.8 for any cancer in families of brain tumor cases and relative risks of brain tumors in these families of 1 to 9. Two casecontrol studies further suggest that risks for other types of cancer may be elevated in family members of brain cancer patients; one found elevated risk of leukemia and liver cancers, and another identified elevated breast and respiratory tract cancer in families. Relatives of children with brain tumors are reported at increased


risk for colon cancer, whereas families with colon polyposis of the colon experience elevated frequencies of gliomas. Also of interest are reports of familial clustering of brain tumors with Hodgkin’s disease.

IV. GENETIC SUSCEPTIBILITY A. Metabolic Susceptibility A refinement of heritability investigations is those into genetic susceptibility, which are genetic alterations that influence oxidative metabolism, carcinogen detoxification, and DNA stability and repair. Genetic technology has described the role of genetic polymorphisms (altered forms of genes established in populations) in modulating susceptibility to carcinogenic exposures. Such a role has been explored in some detail for tobacco-related neoplasms but much less so for other neoplasms, including gliomas. Epidemiology is making widespread use of this technology to examine potentially relevant polymorphisms, including genes involved in carcinogen detoxification, oxidative metabolism, and DNA repair. Earlier studies of a role of genetic susceptibility in the development of brain tumors found variants of cytochrome P450 2D6 (CYP2D6) and glutathione transferase (GSTT1) significantly associated with an increased risk and a significant threefold increased risk for oligodendroglioma associated with the GSTT1 null genotype. The latter study did not find an association between adult onset glioma with either the GSTT1 null genotype or homozygosity for the CYP2D6 variant poor metabolizer genotype. However, when they stratified data by histologic subtype, there was a significant threefold increased risk for oligodendroglioma associated with the GSTT1 null genotype. Negative findings can be fruitful, too, as with Trizna and colleagues, considering the insignificant association of the null genotypes of glutathione transferase , GSTT1, and CYP1A1 and the risk of adult gliomas. The observed pattern of N-acetyltransferase acetylation status proved intriguing, where rapid acetylation produced nearly a twofold increased risk and the intermediate acetylation produced a 30% increased risk. It is unlikely that any single polymorphism will be sufficiently predictive of brain tumor risk. Therefore, a panel of relevant markers integrated with epidemiologic data



should be assessed in a large number of study participants to clarify the role of genetic polymorphisms and brain tumor risk. A more precise estimation of the impact of shared genes, environment, or gene–environment interaction on familial aggregation awaits segregation analysis in population-based studies to test hypotheses of the genetic basis of the aggregation. The sole segregation analysis study of childhood brain tumors rejected the hypothesis that chance explains case families’ cancer patterns; a multifactorial model best explained the data.

B. Mutagen Sensitivity Cytogenetic assays of peripheral blood lymphocytes have been extensively used to determine response to genotoxic agents. The basis for these cytogenetic assays is that genetic damage reflects critical events in carcinogenesis in the affected tissue. To test this hypothesis, Hsu and colleagues developed a mutagen sensitivity assay in which the frequency of in vitro bleomycin-induced breaks in short-term lymphocyte cultures is used to measure genetic susceptibility. This assay was modified using  radiation to induce chromosome breaks because radiation is a risk factor for brain tumors and can produce double-stranded DNA breaks and mutations.  radiation-induced mutagen sensitivity is one of the few significant independent risk factors for brain tumors. DNA repair capability and predisposition to cancer are hallmarks of rare chromosome instability syndromes and are related to differences in radiosensitivity. An in vitro study showed that individuals vary in lymphocyte radiosensitivity, which correlates with DNA repair capacity. Therefore, it is biologically plausible that an increased sensitivity to  radiation results in an increased risk of developing brain tumors because of individuals’ inability to repair radiation damage. However, this finding needs to be tested in a larger study to determine the roles of mutagen sensitivity and radiation exposure in the risk of developing gliomas. The mutagen sensitivity assay has been shown to be an independent risk factor for other cancers, including head and neck and lung, suggesting that the phenotype is constitutional. The breaks are not affected by smoking status or dietary factors (micronutrients).

C. Chromosome Instability A number of chromosomal loci have been reported to play a role in brain tumorigenesis because of the numerous gains and losses in those loci. For example, Bigner and co-workers reported gain of chromosome 7 and loss of chromosome 10 in malignant gliomas and structural abnormalities involving chromosomes 1, 6p, 9p, and 19q; Bello and colleagues reported involvement of chromosome 1 in oligodendrogliomas and meningiomas; and Magnani and colleagues demonstrated involvement of chromosomes 1, 7, 10, and 19 in anaplastic gliomas and glioblastomas. Loss of heterozygosity for loci on chromosome 17p and 11p15 has also been reported. There are little data on chromosomal alterations in the peripheral blood lymphocytes of brain tumor patients. Information on such changes might shed light on premalignant changes that lead to tumor development. We have investigated whether glioma patients have increased chromosomal instability that could account for their increased susceptibility to cancer. Using fluorescent in situ hybridization methods, background instability in these patients was measured at hyperbreakable regions in the genome. Reports indicate that the human heterochromatin regions are frequently involved in stable chromosome rearrangements. Smith and Grosovsky (1993) and Grosovsky and co-workers reported that breakage affecting the centromeric and pericentromeric heterochromatin regions of human chromosomes can lead to mutations and chromosomal rearrangements and increase genomic instability. Our study demonstrated that individuals with a significantly higher level of background chromosomal instability have a 15-fold increased risk of development of gliomas. A significantly higher level of hyperdiploidy was also detected. Chromosome instability leading to aneuploidy has been observed in many cancer types. Although previous studies have demonstrated the presence of chromosomal instability in brain tumor tissues, our study was the first study to investigate the role of background chromosomal instability in the peripheral blood lymphocytes of patients with gliomas. This suggests that accumulated chromosomal damage in peripheral blood lymphocytes may be an important biomarker for identifying individuals at risk of developing gliomas.


V. ETIOLOGY AND RISK FACTORS A. Ionizing Radiation Research consensus holds that therapeutic, not diagnostic, ionizing radiation is a strong risk factor for intracranial tumors. Even relatively low doses (averaging 1.5 Gy) for ringworm of the scalp (tinea capitis) have been associated with relative risks of 18, 10, and 3 for nerve sheath tumors, meningiomas, and gliomas, respectively. Other studies report a prevalence of prior therapeutic radiation among (17%) patients with glioblastoma or glioma and increased risk of brain tumors in children after radiation for acute lymphoblastic leukemia. Shorter term, diagnostic radiation apparently plays no role; three case-control studies of exposure to dental X rays reported relative risks of 0.4, 1.2, and 3.0 for gliomas. For meningioma, three of four studies reported risks higher than 2, but because the results all issue from one site, they invite authentication from other geographic areas. Risk due to prenatal or occupational radiation exposure remains unknown, as studies of in utero exposure to atomic bomb or occupational radiation have found standard brain tumor incidence or results that are confounded or contradictory. Small-scale prenatal exposure studies are statistically insignificant, and the slightly elevated risk associated with a comparatively uncommon radiation exposure would not explain most childhood brain tumors. The study finding a small but statistically significant elevated risk of 1.2 for brain tumors in nuclear facility employees and nuclear materials production workers allows for the possibility of confounding or effect modification by chemical exposures. A study reporting increased mortality from brain tumors among airline pilots, possibly implicating exposure to cosmic radiation at high altitudes, is contradicted by another study reporting standard morbidity rates in pilots.

B. Electromagnetic Fields Despite largely negative findings, the debate on the impact of electromagnetic fields on brain cancer continues, prolonged by methodological difficulties with some studies and popular media warnings about consumer products inviting personal exposures. In 1979,


Wertheimer and Leeper described an apparent increased risk of brain tumors and leukemia in Denver children living near high-current versus low-current wiring. Their report set off widespread public and scientific interest in the potential health effects of electromagnetic fields and electronic devices, but positive findings came into question after contrary outcomes and reviews of sample sizes and method. Among the studies reducing suspicion about electricity was a meta-analysis showing an insignificant increased childhood brain tumor risk for residents in homes coded for high current. In a meta-analysis of 29 studies of adult brain tumors in relation to occupational exposures to electric and magnetic fields, apparently adding to the weight of evidence against electricity, Kheifets and colleagues reported a significant (10 to 20%) increased risk for brain cancer among electrical workers, but could not show a consistent dose–response relationship or different outcomes for electrical workers at different exposure levels. Nor could risk be shown in large population or epidemiologic studies of the relation of brain tumor to EMF. In a San Francisco study, 492 adults with glioma and 463 controls were equally likely to have lived in homes with high wire codes during the 7 years before diagnosis. Spot measurements in the homes also showed that cases and controls experienced the same level of electromagnetic fields at home. EMF investigation continues in the quest of reliable risk determination, but exposure research reveals inconsistencies. Electromagnetic field measurement in the home is inexact because of varying wire codes and the spot measurement snapshots usually taken for studies and, as well, the neglect of long-term exposure measurement or factoring in additional potential exposures from internal wiring, appliances, and gadgets. Temporal, intensity, and external influences may overshadow domestic exposures. The complexity required by further investigation is suggested by a Swedish study reporting increased risks of adult leukemia and central nervous system tumors after both residential and occupational exposure but not after either separately. The study used detailed EMF information preserved over decades from Swedish power suppliers recording levels of exposure. Such information is unavailable in the United States, but, again, the study lacked specificity about the exact daily experience of



its subjects. The difficulties of quantification in EMF research pale compared to the main problem in this field of investigation: electromagnetic fields are incapable of inducing mutations that in turn might promote tumorigenesis. The EMF studies, as well as several occupational or environmental reports and many cluster reports, lack the criteria for causal relationships proposed by Bradford Hill, among which are specificity, consistency, and plausibility

C. Diet Studies of the susceptibility of primates and other mammals to chemically induced brain tumors have prompted extensive study of diet and brain tumors. Experimental animal studies find N-nitroso compounds are clearly neurocarcinogens; other investigations describe mechanisms involving DNA damage through which N-nitroso compounds might cause brain tumors. These compounds can initiate neurocarcinogenesis through both prenatal and postnatal exposure, although in animals, more tumors result from fetal than postnatal exposures. The lag reported between exposure and tumor formation opens the possibility of the implication of early exposure in adult tumors. However, the ubiquity of N-nitroso compounds complicates human dietary studies, nonetheless eagerly pursued, especially into the salutary effects of vitamins. N-nitroso compounds arise in the digestive system in a promoting enzymatic milieu when common amino compounds from fish, other foods, or drugs contact a nitrosating agent, such as nitrites in cured meats. Some vegetables also contain nitrates convertible to nitrites but, as well, contain vitamins that block the formation of N-nitroso compounds. Despite great effort, human diet studies have achieved no consensus and provide only limited support for their motivating hypothesis, that dietary N-nitroso compounds heighten the risk of both childhood and adult brain tumors. However, some studies find that brain tumor cases or their mothers consumed more nitrosamines than controls, whereas others note lower tumor rates after a comparatively elevated consumption of fruits and vegetables or vitamins that might block nitrosation or harmful effects of nitrosamines. Ambiguities are evident. For example, a case con-

trol study by Lee and co-workers found that adults with glioma were more likely than controls to consume diets high in cured foods or nitrites and low in vitamin C, but the effect only achieved statistical significance in men. The finding is compatible with the hypothesis that N-nitroso compounds play a role in human neurooncogenesis, but the observed patterns also support the hypothesis of oxidative burden and antioxidant protection. Also, despite much searching, particularly in connection with pediatric brain tumors, neither alcohol nor tobacco has been clearly implicated in brain tumors. Their demonstrable impacts, especially fetal, require concern. Beer and malt liquors alone among alcoholic beverages are suspect due to the oxidation of malt, a nitrosamine derivative. Despite its other dangers and the polycyclic aromatic hydrocarbons and nitroso compounds in its smoke, tobacco has not been established as a causative agent in primary brain tumors, although two studies claim increased glioma risk lies in unfiltered cigarettes. The risk for secondhand smoke seems more clearly established; reports support an elevated risk of near 1.5, roughly equal to the elevated risks in passive smokers for adult lung cancer or cardiovascular disease. Most frequently associated with maternal smoking or passive smoke exposure are childhood brain tumors and leukemia– lymphoma, with risks going up to two or more in selected studies. Paternal as well as maternal smoking produces risk, suggest some. Even in the absence of definitive findings in human studies, the demonstrations of fetal genotoxicity from metabolites of tobacco smoke, and the demonstrable presence of adducts, should lead to strong recommendations for mothers to reduce fetal and infant exposure to tobacco smoke.

D. Industry and Occupation Attempts to link specific chemicals to human brain tumors in specific work groups, hampered by small numbers and the difficulties of segregating one of the many agents in the workplace, as well as inherent problems in retrospective studies, have proven inconclusive. Findings from animal experiments may be inapplicable or irreproducible in humans. Chemical implantation shown to provoke tumors in animals differs


from inhalation or dermal exposures in occupational settings. Nor are factors clearly altering animal susceptibility such as strain, gestational age, and fetal versus adult status useful for occupational studies. Follow-up studies of occupationally induced brain cancer usually consist of too few affected subjects to establish or pinpoint causal chemicals, physical agents, work processes, or interactions. Thus, suspect chemicals and occupations listed in 1986 by Thomas and Waxweiler, despite many more recent studies, retain their ambiguous status, although among them are agents with undoubted animal neurocarcinogenicity, such as organic solvents, lubricating oils, acrylonitrile, formaldehyde, polycyclic aromatic hydrocarbons, phenols, and phenolic compounds. Cancer clusters, forming no risk pattern, have been found in chemical, automotive, and textile industries. Possibly the strongest case can be made against vinyl chloride, lethal in rats, and in 9 of 11 studies of human workers associated with as much as a twofold increased relative risk of dying of brain tumors. Animal studies usually involve implantation of a single chemical rather than skin or lung exposures to many substances at the same time that workers experience. Thus, because of the problem, no definitive link has been established between brain tumors and even strongly suspected carcinogens. For example, organochlorides, alkyl ureas, and copper sulfates compounds that induce cancer in laboratory animals only randomly produce risk in agricultural workers. Study design faults and small numbers plague the occupational studies, so far unable to agree on risks involved in manufacturing pesticides or fertilizers, but reporting in four of five studies of pesticide applicators a nearly threefold risk elevation. Because they involve the production of many suspect carcinogens, synthetic rubber production and processing have received careful scrutiny by investigators who generally found a median increase in brain tumors of as much as 90%. The by-products of synthetic rubber processing, such as coal tars, carbon tetrachloride, N-nitroso compounds, and carbon disulfide, might appear to account for this increased risk of brain tumors. However, several studies showed no increased risk or a decreased risk of brain tumors in this industry, and studies have usually failed to show a link with a single chemical.


With formaldehyde, another long-suspected compound, the numbers are greater but conclusions are similarly elusive. Formaldehyde, to which nearly two million U.S. workers are exposed, causes cancer in laboratory animals. However, Blair and co-workers, in evaluation of 30 epidemiological studies of segments of this large group, concluded that the risk was 50% elevated for those exposed in professional roles such as embalmers, pathologists, and anatomists but not for industrial workers with formaldehyde exposure. Blair and colleagues therefore rejected a causal role for formaldehyde in human brain tumorigenesis, a finding with impact in the workplace. Other unknown cofactors may obscure the true risk in industrially exposed workers and create a skewed estimate of risk in occupational groups.

E. Viruses Certain viruses, like the suspect chemicals, have been found to induce brain tumors in animal studies. As in chemical studies, small numbers and negative findings hinder epidemiological evaluation. Repeatedly, calls have been made for aggressive studies addressing the role of viruses (and other infectious agents) in causing human brain tumors. Unfortunately, very few epidemiological studies have addressed the virus–tumor relationship, probably because of the difficulties in designing meaningful studies. Viruses and infectious agents could be an explanation for a proportion of brain tumors, and therefore intriguing as virology advances. Contamination by the simian virus of the widely distributed Salk vaccine for polio offered the numbers from which to derive significant statistics, as 92 million United States residents received it. However, as with other risk exposure studies, investigations of SV40 were flawed, often anecdotal, based on questionable recall, or amenable to confounding factors. They offered hints, clues, or perhaps merely coincidences. Generally no association between virus and cancer could be established, although it may not be ruled out because the level of contamination varied among lots and manufacturers and was not taken into account in the United States and in Germany. Association was found small to none in a variety of studies, including a cohort study of the risk of



ependymoma, osteosarcoma, and mesothelioma and a rare case-control study in England. Maternal contaminated vaccination risks for childhood cancer and brain cancer in particular are from sketchy and possibly confounded reports. Viral disease investigations for protective or heightened risk effects are similarly contradictory. After reports that mothers of children with medulloblastoma were exposed in pregnancy to chicken pox, a herpes virus, Wrensch and colleagues found that mothers of glioma cases had lower rates than controls of chicken pox or shingles. This observation was supported by serologic evidence that cases were less likely than controls to have antibody to varicella zoster virus, the agent for chicken pox and shingles. There is some plausibility that viruses and infectious agents could be an explanation for a proportion of brain tumors, and therefore intriguing as they are still in their earliest stages. The JC virus has come under scrutiny because it is excreted in the urine of immunosuppressed, immunodeficient, and pregnant women and was found associated with medulloblastoma in children and other tumors. However, the JC virus exists in cancer-free subjects and its connection, if any, to tumorigenesis is only a surmise.

F. Drugs and Medications Preliminary studies have generally reported an insignificant association between tumors and medications, including pain, headache, sleep, fertility drugs, oral contraceptives, tranquilizers, antihistamines, and diuretics. Ryan and colleagues found that diuretics have a nonsignificant protective association against meningioma, which is opposite for adult glioma. They also found little association between antihistamine use and adult glioma, but a 60% increased relative risk for meningioma. Prenatal exposure to diuretics was half as common among children with brain tumors than among controls in two studies, but twice as common in one study. Prenatal exposure to barbiturates has not been consistently or convincingly linked to childhood brain tumors. As nonsteroidal antiinflammatory drugs may be protective against certain cancers, the role of these drugs in brain tumors should be investigated.

G. Cellular Telephones The use of cellular telephones has grown remarkably over the last decade, and it is estimated that more than 500 million individuals worldwide use handheld cellular devices. The telephones contain a small transmitter that emits radio frequency radiation next to the head, and there has been great public concern that individuals exposed to radiation emitted from wireless communication technologies might have an increased risk of developing tumors of the brain and nervous system. To date, six papers have been published, none of the studies support the hypothesis of an association between use of these telephones and tumors of the brain or other cancers. Rothman and colleagues reviewed mortality among more than 250,000 customers of a large cellular phone operator in the United States and did not find an increased risk after a follow-up of only 1 year. The numbers of brain cancers (n  6) and of leukemias (n  15) were small, and there were no statistically significant associations with number of minutes of phone use per day or years of phone ownership. The second study, a case-control study from Sweden by Hardell, reported a statistically nonsignificant increased risk for brain tumors on the side of the head on which cellular telephones were used. However, the risk for brain tumors overall was not increased, and there were methodologic concerns related to the ascertainment of cases. The fourth report was a case-control study from five academic institutions in the United States with 469 patients with primary brain cancer and 422 matched controls between 18 and 80 years of age. They found no association with brain cancer by duration of use (p  0.54). In some cases, cerebral tumors occurred more frequently on the same side of the head where cellular telephones had been used (26 vs 15 cases; p  0.06), but in cases with temporal lobe cancer, a greater proportion of tumors occurred in the contralateral than ipsilateral side (9 vs 5 cases; p  0.33). The fifth study was also a hospital-based casecontrol conducted by investigators at the National Cancer Institute. They included 782 patients and 799 controls. The relative risks associated with cellular phone use for more than 100 h was 0.9 for gliomas [95% confidence interval (CI)  0.5–1.6] and 0.7 for meningioma (95% CI  0.3–1.7); 1.4 for acoustic


neuroma (95% CI  0.6–3.0). They found no evidence that the risks were higher among persons who used cellular phones for more than 5 years, nor did they observe more tumors on the side of the head the phone was typically used. The sixth study was a retrospective cohort study of cancer incidence conducted in Denmark using subscriber lists from the two Danish operating companies. They identified 420,095 cellular telephone users during the period from 1982 through 1995 and linked the list with the Danish Cancer Registry. They observed 3391 cancers overall and expected 3825 cases (SIR  0.89; 95% CI  0.86 to 0.92); the risk cancers of the brain or nervous system were also lower than expected (SIR  0.95; 95% CI  0.81–1.12). In addition, a large occupational cohort mortality study among 195,775 employees of Motorola, a manufacturer of wireless communication products, did not support an association between occupational radio frequency exposure and brain cancers or lymphoma/leukemia. To date, the studies all seem to support the hypothesis that there is no association between use of these telephones and tumors of the brain or other cancers.

VI. CONCLUSIONS The most generally accepted model, and productive of the most fruitful research, of carcinogenesis holds that cancers develop through the accumulation of genetic alterations allowing the cells to escape regulatory mechanisms and/or destruction by the immune system. Some inherited alterations in crucial cell cycle control genes, such as p53, as well as chemical, physical, and biologic agents that damage DNA, are therefore considered candidate carcinogens. Although rapid advances in molecular biology, genetics, and virology promise to help elucidate the molecular causes of brain tumors, continued epidemiologic work will be necessary to clarify the relative roles of different mechanisms in the full scope of human brain tumors. Genetic and familial factors implicated in brain tumors have been the subject of many studies. In summary, the etiology of brain tumors remains largely unknown. We now know that primary brain


tumors have many causes. Because not one cause thus far identified accounts for a very large proportion of cases, many possibilities remain that will enable us to discover important risk factors. Moreover, in the continuing search for explanations for this devastating disease, new concepts about neurooncogenesis might emerge, making the study of brain tumor epidemiology particularly exciting. The family studies conducted thus far suggest some role for inherited susceptibility to CNS tumors. Until the gene or genes for brain tumors are identified, genetic counseling in families at high risk of brain tumors is not possible. However, individuals with specific hereditary syndromes that predispose to brain tumors can be told their genetic risks.

Acknowledgments This work was supported in part by Grants RO1-CA52689 and PO1-CA55261A from the National Cancer Institute.


Bibliography Berleur, M. P., and Cordier, S. (1995). The role of chemical, physical, or viral exposures and health factors in neurocarcinogenesis: Implications for epidemiologic studies of brain tumors. Cancer Causes Control 6, 240–256. Blatt, J., Jaffee, R., Deutsch, M., et al. (1986). Neurofibromatosis and childhood tumors. Cancer 57, 1225. Bondy, M. L., Lustbader, E. D., Buffler, P. A., et al. (1991). Genetic epidemiology of childhood brain tumors. Genet. Epidemiol. 8, 253. Bondy, M., Wiencke, J., Wrensch, M., et al. (1994). Genetics of primary brain tumors: A review. J. Neuro-Oncol. 18, 69–81. Choi, N. W., Schuman, L. M., and Gullen, W. H. (1970). Epidemiology of primary central nervous system neoplasms. II. Case-control study. Am. J. Epidemiol. 91, 467. Draper, G. J., Heaf, M. M., and Kinnier Wilson, L. M. (1977). Occurrence of childhood cancers among sibs and estimation of familial risks. J. Med. Genet. 14, 81. Grosovsky, A. J., Parks, K. K., Giver, C. R., and Nelson, S. L. (1996). Clonal analysis of delayed karyotypic abnormalities and gene mutations in radiation-induced genetic instability. Mol. Cell. Biol. 16, 6242–6262.



Kleihues P, Cavenee W. K. (eds.) (1997). Tumors of the central nervous system: pathology and genetics. International Agency for Research on Cancer, Lyon, France. Kuijten, R. R., Bunin, G. R., Nass, C. C., et al. (1990). Gestational and familial risk factors for childhood astrocytoma: Results of a case-control study. Cancer Res. 50, 2608. Li, F. P., and Fraumeni, J. F., Jr. (1975). Soft tissue, breast cancer and other neoplasms. Ann. Int. Med. 83, 833. Lynch, H. T., Guirgis, H. A., Lynch, P. M., et al. (1977). Familial cancer syndromes: A survey. Cancer 39, 1867. Magnani, C., Pastore, G., Luzzatto, L., Carli, M., Lubrano, P., and Terracini, B. (1989) Risk factors for soft tissue sarcomas in childhood cancer: A case-control study. Tumori 75, 396–400. Muscat, J. E., Malkin, M. G., et al. (2000). Handheld cellular telephone use and risk of brain cancer. JAMA 284, 3001–3007. Preston-Martin, S., and Mack, W. J. (1996). Neoplasms of the nervous system. In “Cancer Epidemiology and Prevention” (D. Schottenfeld and J. F. Fraumeni, eds.), 2nd Ed., pp. 1231–1281. Oxford Univ. Press, New York.

Rothman, K. J., Loughlin, J. E., Funch, D. P., and Dreyer, N. A. (1996). Overall mortality of cellular telephone customers. Epidemiology 7, 303–305. Ryan, P., Lee, M. W., North, J. B., et al. (1992). Risk factors for tumors of the brain and meninges: Results from the adelaide adult brain tumor study. Int. J. Cancer 51, 20–27. Schwarz, B., and Schmeiser-Rieder, A. (1996). Epidemiology of health problems caused by passive smoking. Wien. Klin. Wochenschr. 108, 565–569. Seizinger, B. R., Rouleau, G. A., Ozeluis, L. J., et al. (1987). Genetic linkage of von Reckinghausen neurofibromatosis to the nerve growth factor receptor gene. Cell 49, 589. Weinberg, R. A. (1991). Oncogenes, tumor suppressor genes, and cell transformation: Trying to put it all together. In “Origins of Human Cancer” (J. Brugge, T. Curran, E. Harlow, et al., eds.), Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Wrensch, M., Bondy, M. L., Wiencke, J., et al. (1993). Environmental risk factors for primary malignant brain tumors: A review. J. Neurooncol. 17, 47–64.