Document 2RyyjJkJ5307mL7ZnwmrDvq7L

CANCER EPIDEMIOLOGY and PREVENTION DAVID SCHOTTENFELD, M.D. Chief of Epidemiology and PreventiveMedicine Memorial Sloan-KetteringCancer Center New York, New York JOSEPH E FRAUMENI, JR., M.D. Chief, Environmental Epidemiology Branch National Cancer Institute National Institutesof Health Bethesda, Maryland 1982 W. B. SAUNDERS OMPANY Philadelphia London Toronto Mexico City Rio de Janeiro Sydney Tokyo e CHAPTER FORTY-ONE The Leukemias Clark W. Heath, Jr. 4 - i INTRODUCTION The leukemias encompass a diverse group of nialignancies that share in common merely the fact that all arise from cell systems that circulate in peripheral blood and that arise in large part from bone marrow. The differing cell origins of the leukemias require that different leukemic cell types be considered separately i r i terms o f c.tiology, pathogenesis, therapy, and prevention. The convenience o f the collective tern1 "leukemia" frequently results i n the leukemias bring considered as a single entity. For general purposes, at least three separate types of leukemia should be distinguished: acute leukemia (AL), chronic granulocytic leukemia (CGL), and ( hronic lymphocytic leukemia (CLL). Taken as a whole, the leukemias are not a particularly comrrion form of human cancer, representing only about 5 per cent of total annual cancer incidence (Cutler and Young, 1975; Young et al, 1978). Despite being uncommon, however, the leukemias over ,the years have been studied more extensively than most other cancers. In part, this reflects the fact that these types of malignancy involve tissue systems (blood and marrow) that are easily a(-(wsible for diagnosis, for studies of pathogenetic mechanisms, and for monitoring therapeutic effects. In fact, not only are leukwiic cells especially accessible for sampling, but thvy can be separated more readily from other t i w a components than is possible for cancers involving solid tissues. The ability to examine and study relatively pure preparations of malignant cells i s a distinct advantage for research st i r c l ips. Research concerning leukemia has also been greatly facilitated, in contrast to many more 720 common human cancers, by the existence of several readily available animal models. As a consequence of these models and their availability for research, extensive information, espe- cially with respect to etiologic mechanisms, has come from studies of leukemias as they occur in murine, avian, feline, bovine, and simian s p ' - cies. Despite these practical advantages for re- search work, our understanding of leukeniia etiology and pathogenesis has yet to produce effective methods for primary disease prevention, except perhaps for minimizing man-made exposures to ionizing radiation and to such marrow-toxic chemicals as benzene. This state of affairs stands in striking contrast to our ohvious capacify, long recognized, for preventing most lung cancers through curtailment of ciga- rette smoking. This disparity in capacity for primary prevention unclerscores what may tie a fundamental difference between tumors of mesothelial or mesenchymal origin (of which the leukemias are one class) and tumors of epithelial origin (such as lung cancer). For epithelial types of cancer, it appears that environmental factors arise more generally from features of personal lifestyle (for example, diet, cigarette smoking, alcohol use, sexual behavior, and s o on), while for nonepithelial cancers such etiologic influences appear to be less important. To the extent that we are capable of modifying lifestyle factors, this difference may be translat- able into a greater capacity for prevention of epithelial tumors despite greater insight into the etiology and pathogenesis of mesothelial tumors such as the leukemias. The capacity to prevent disease by no means requires full understanding of disease causat'ion. This chapter contains a review of current Tk1E LEUKEMIAS -729 knowledge concerning the distribution and causes of human leukemias, with particular attention given to the implications such knowledge niay have for preventive strategies. Since most of what we know concerning the etiology of human leukemias comes from epidemiologic studies, bolstered b y experimental work in lower species, the emphasis here i s w :lie tpitleniiology of leukemias. Since an earlier rcwiew summarized the medical literature published through 1973 (Heath, 19751, the present survey is designed t o summarize current knowl- edge, with emphasis on work pihlislied more recentIy . HISTORICAL PERSPECTIVE Kelative stanrlardization and specific ity of disease c Iassification codes in the 19305, coupled with improved diagnost ic capaI 1iI i ties through expanded use of bone marrow aspiralion in the 1940s, made effective epicleniiologic \tuditts of leukemia pnssihle for the first time in the yeais following World War It. Prior to that tiriie, etiologic observations consisted largely of c Iinic at anecdotes, with considerable confusion pxisting between leukemias and infectionassociated leukocytosis. Perhaps because of su- perficial diagnostic similarities between leuke- mia and leukocytosis, early theories of leukemia f.tiology heavily emphasized the possible role of infectious agents. As the medical uses of ioniz- ing radiation increased, however, and as cases of lcukeniias l q a n to appear i n persons involved i n such use (radiologists), radiation received greater attention as a cause of leukemia. This recognition, of course, was dramatically reinforced i n the early 1950s when increased leukemia incidence first began to appear among Japanesesurvivors of atomic bomb radiation. At the same time, great attention began to focus on the viral origins of aninial cancers, partit d a i l y leukemia in mice. Such work recapitulated studies performed in the early 1900s tfiat c learly identified the viral origin of chicken leukosis (Heath et al, 1975). Renewed animal exywrinients in t h i s field were made possible by the development of inbred strains of mice, srveral of which showed marked susceptibility t o leukemia. With the availability of such a ( onvenient experimental tool, rapid advances were made in identifying animal leukemia viruses. In this setting, attention focused sharply during the 1960s on the possible viral origins of human leukemias. This was accompanied by increased interest in the possible genetic origins of leukemia, partly since animal models sug- gested that vertical transmission of leukemia virus from generation to generation might be an underlying mechanism for familial tendencies in the disease. Identification of human leukemia viruses, however, has proven to he an elusive task, and so, in the past decade, etiologic theories rcgartting leukemogenesis have givcn more and niorc' recognition to multiple causation, with special emphasis on environmental causes, particularly possible chemical exposures. The latter trend follows closely the recent tendency to emphasize theories of environmental carcinogenesis i n the entire cancer field. A broad historic view of etiologic thinking with regard to the leukcmias, however, would suggest that, both for these particular types of cancer and for cancers as a whole, a balance of interactive causative factors, some arising in the environment and some in the host itself, rather than any one form of causation, comprises the proper perspective in studies of cancer etiology. PATHOLOGIC CHARACTERISTICS As emphasized initially, the leukemias constitute a collection of quite different diseases. Cases are basically classified according to cytology of leukemic cells coupled with certain histologic, cytochemical, and immunologic characteristics. Much of the clinical and pathologic variation described among the leukeniiaq, however, either i s primarily of concern for predicting prognosis and response to therapy or is of rather uncertain significance. For the purpose of understanding current epidemiologic data, it seems sufficient to recognize three basic types of leukemia (AL, CGL, and CLL) and to be aware of the fact that often the lymphocytic varieties of leukemia are closely related to lymphomas. In the latter circumstance, occasional cases, especially cases of CLL, either may have progressed from what previously had been diagnosed as lymphoma or niay differ clinically from lymphoma only by the fact that bone marrow i s found to be infiltrated with lymphocytes. This diagnostic overlap between lympliomas and lymphocytic leukemias suggests that for some varieties of leukemia, considerations of epldemiology and disease pievention should involve both categories of cancer togetliw. At the same time, certain rare forms of leukemia (plasma cell leukemia, for instance) might for epidemiologic purposes more appropriately be considered in other diagnostic categories (multiple myeloma). Of the three basic forms of leukemia men- 'V 4 r1 s: 5$"r' 730 - CANCER BY TISSUE OF ORIGIN tioned, the grea' ?st clinicopathologic variety i s encountered for AL. Here a wide range of cell types can be seen. The variety is generally greater in adult than in childhood cases, acute lymphocytic leukemias (ALL) and acute undifferentiated (stem cell) leukemia accounting for the great majority of childhood cases. Overall, acute granulocytic leukemia (AGL), ALL, and acute undifferentiated leukemias are the varieties of AL most commonly cited. Two not uncommon cytologic varieties, however, especially among adults, are acute monocytic leukemia and acute myelomonocytic leukemia. The latter form in particular is considered by many hematologists to be a variety of AGL. True acute monocytic leukemia, on the other hand, may have its origins in the monocyte or histiocyte rather than in the granulocytic cell series. Commonly, cases of granulocytic leukemia may exhibit disordered red cell proliferation in niarrow, with occasional abnormal nucleated red cells appearing in peripheral blood. Such cases, especially when the red cell abnormalities are pronounced, are commonly termed erythroleukemia or diGuglielmo's disease or syndrome. Whatever terminology is used, however, clinically and epidemiologically such cases seem likely t o belong under the classification of ACL. Cytogenetic studies have been widely used in qtudies of leukemia cell variations. Despite c"xtensive investigations, however, the nnly cytogenetic finding of consistent value in distinguishing among cases has been the Philadelphia (Ph') chromosome, which characteri7es marrow cells in most cases of typical CGL. Occasional cases of CGL that do not display this cytogenetic characteristic appear to IF a distinctly separate pathologic group with different clinical characteristics and different responses to therapy. Cases of CGL ambng children, although not a common occurrence, are more usually of the Ph'-negative variety, whereas the reverse holds true in adult cases. EPIDEMIOLOGIC CHARACTERISTICS The descriptive epidemiology of the leukem i a s can be considered in terms of (1) secular frcnds, (2) geographic distribution, and (3) age, sex, and race or ethnic group patterns. Although detailed cancer incidence data have recently become more widely available (Cutler and Young, 1975; Waterhouse et al, 1976; Young et al, 19781, much of our knowledge concerning the time, place, and person distribution of leu- kemias still depends on mortality statistics, While such data can provide a fairly accurate reflection of incidence for acute fornis of leukemia, variable survival times and inaccuracies in death registration limit the value of mortality information. Secular Trends Leukemia incidence, as reflected in mortality data, increased dramatically from 1900 to the 1940s. While much of this rise undoubtedly reflected improved diagnosis and more accurate death registration, some at least can be presumed to reflect true increased occurrence, perhaps the result of increased exposure to radiation and chemicals in industrial socicties. Since the 1940s the rate of increase has diminished, with most recent data suggesting a slight decline in rates, especially among whites. While this trend may well reflect improved control over leukemogenic environmental exposures i n the last several decades, it might also be attributed to lengthening case survival coincident with attainment of relatively complete case detection and reporting. Geographic Distribution For cancers overall, striking differences in frequency exist between different countries. This is considerably less true for the leukemias, however, t h h for most other forms of malignancy. This fact may well reflect underlying etiologic differences between epithelial and mesothelial tumors, whereby striking variations in lifestyle factors are mirrored in striking variations in incidence levels for epithelial cancers. For the leukemias, however, geographic differences rarely exceed the twofold level for coun- tries with comparable reliability and accuracy in the reporting of mortality or incidence statis- tics. The most striking difference involves CLL, which is extraordinarily uncommon in Oriental countries. Since this difference in CLL frequency persists for persons of Oriental origin living in other countries, it seems likely that host or genetic factors are responsible (Heath, 1976). For other types of leukemia, differences in incidence are much less marked, and it is uncer- tain to what extent they reflect environmental as opposed to host factor variations. Small variations in incidence levels have been described for regions within countries, notably the United States and Great Britain. THE LEUKEMIAS -731 Wliilc these differences rarelFkxceed a 1.5-fold differential, they do appear to correlate with socioeconomic conditions, higher rates existing where income and physician density is higher. This is also reflected in a slight tendency for r a t m to be higher in urban as opposed to rural scttings. Despite these findings, however, the overall picture is one of relative uniformity, cspecially when compared to geographic variations in other cancer types. increasing male to female ratio over time, such a trend is not clear-cut. Since excess male incidence is commonly equated with possible occupational factors operating preferentially by sex, this general pattern for the leukemias may possibly reflect the influence of occupational environmental leukemogens. Race and Ethnic Group Age As with all cancers, the leukemias vary widely in incidence by age. Overall annual in( iclcnce rises with age from a low of 1 to 2 cases pw 100,000 in young adulthood to 30 or more rases per 100,000 over age 50. Unlike most other canc ers, however, a distinct childhood pcak exists, rising to about 5 cases per 100,000 at age 2 t o 4 years and decreasing in later ( I i i l ~ l l i o o c l T h i s peak tias appeared especially t o 'iffm t Western white populations, a n d in Great Britain i t has been suggested that the childhood pcak may not have existed prior to the 1920s. W h k the cause of the peak i s debatable, it is conc eivable that i t reflects changing environnicntal exposures in utero or in early infancy. Different leukemic cell types display striking diffrwnce5 in age pattern. CLL i s almost never seen tx3fore niid-adulthood and is coninion a h v e age 50. Ph'-positive CGL is likewise an adult disease, although not uncommon in younger adults, and is occasionally even seen in children. As indicated previously, when CGL does appear in children, it more often is of a Ph'megative variety. At all ages, the most common variety of leukemia i s AL, accounting for virtually all childhood and young adult cases and perhaps two-thirds of cases in older adults. In childhood, most cases of AL appear as ALL or undifferentiated or stem cell types. After puberty the great majority of AL cases, however, are AGI or a monocytic or myelomonocytic variety. 1he origins of these age differences are unknown. Sex Males in general are affected more often by leukemia than are females. Sex ratios, however, are to some extent a function of cell type: male predominance i s greatest for CLL @ : I ) , while the sex distribution is closer to 1:l for other types. Although some data have suggested an Racial and ethnic patterns of leukemia incidence show a consistent excess at all ages in whites over blacks and among Jewish populations. To what extent these differences reflect host or genetic factors as opposed to socioeconomic or environmental and lifestyle influences i s not known. Among blacks, however, the leukemia deficit is particularly prominent in older age groups and may to some extent represent deficient case detection as a tunc-tion of medical care availability. In contrast, the striking deficit of CLL i n Oriental populations, as discussed previously, appears t o arise priniarily from genetic factors or some intrinsic racerelated host characteristic governing leukemia susceptibility. ETIOLOGIC EVIDENCE As i s true of cancer generally, the probable causes of leukemia include a wide range of factors, acticg singly or in combinations, some involving intrinsic host characteristics (genetic traits and immunologic mechanisms), and others reflecting environmental exposures (radiation, chemicals, and infectious agents). Information comes both from animal experiments and from epidemiologic studies of human populations. What follows is an overall synopsis of epidemiologic data, with emphasis on observations published since 1973. Genetic Factors Various lines of evidence indicate the operation of host genetic factors in human leukemia etiology. Aside from differences among racial groups (especially the paucity of CLL among Oriental populations as noted previously), such evidence includes multiple cases in families and associations between leukemia occurrence and various genetic markers or genetically determined conditions (Heath, 1976; Zuelzer and Cox, 1969). 4 d 732 - - C'ANCER BY TISSUE OF ORIGIN Familial f eukemia Although occasional families have exhibited striking tendencies for multiple leukemia case occurrence (Kaur et al, 19721, such situations are milch more the exception than the rule. Numerous surveys have examined the question of familial leukemia occurrence without consislent evidence that leukemia incidence is more than faintly increased in close relatives of leukem i a patients. If a general tendency does exist, i t :vould appear to involve CLL more often than other cell types. The existence of consanguinity in occasional multiple case families, however, together with an apparent tendency for increased consanguinity in Japanesefamilies with lciikemic s i h pairs (Kurita and Kaniei, 1969), sttongly suggests that genetic factors can be important in human leukemia, much like aninial counterparts in which experimental inheeding can dramatically affect frequency of Icukeniias and other cancers. Particular attention has been given to twin studies with somewhat uncertain results. Al- though some observations in the United States have suggested strong concordance of childhood leukemia in like-sexed or monozygotic twins, evidence from Great Britain i s uncertain (Draper et at, 1977). Given the fact that mono7ygotic twins share their placental circulation, it itlay be that some cases of concordant twin Ictrkemia represent intrauterine leukemia "transplantation" from an affected twin to its mate rather than genetic predisposition alone (Chaganti et al, 1979; Falletta et al, 1473). Associated Genetic Conditions More striking evidence, perhaps, of a genetic Lit ior in leukemia etiology involves association of leiikemia cases with'chromosomal aberra- tions, increased incidence among individuals with known chromosomal disorders, and in- r t w e d incidence in connection with certain itnniunologic diseases. Despite such clinical c w o c iations, no clear-cut or consistent genetic tnarkers (e.g., A B 0 blood group and HLA triztocompatibility type4 have yet been discovw ~inl connection with leukemia (Blattner et al, 1078). ('hroniosomal abnormalities, however, an in- tlircc t indication perhaps of the fundamental tc4.ition of leukemia to the structure and func- tion of genetic material, are commonly seen in Icukeniic bone marrow cell lines. These aberrations (increased chromosomal breakage and ahnornmal karyotype patterns) vary widely from case t o case. Tn date, the only chromosomal abnormality specific to a particular leukemic cell type is the Phl chromosome in niarrow cells from cases o f CCL. Beyond such chromosomal changes in incli- vidual cases, leukemia has long been known to occur with greatly increased frequency in Down syndrome (G trisomy) and probably in other varieties of nieiotic nondisjunction (D1315 trisomy and Klinefelter syndrome). Among persons with Down syndrome in particular, leukemia i s about 20 times more common than in the general population. All ages are apparently affected, and all types of leukemia are probably involved. A further parallel between these two disorders i s the fact that both Down syn- drome and leukemia (at least among children) increase in frequency with mother's age at time of birth. These relationships, however, probahly operate independently of each other since, unlike Down syndrome, leukemias appear t o decrease in frequency with increasing birth order. A further association i s the apparent increased frequency of Oown syndrome among sibs of leukeniic children. In addition to chromosomal aberrations in leukemia cases a n d the association Ix>tween leukemia and nondisjunction syndronwc, a fur- ther link to chromosomal disturbance is the fact that leukemias occur with increased frequency in various conditions themselves charncterized b y increased chromosomal breakage. Several of these conditions are hereditary in nature: Bloom syndrome and Fanconi anemia. Others rcfkct environmental exposures that induce chromosomal injury: populations exposed to ionizing radiation or to chemicals such as Iwn- zene. In fact, the one form of leukemia consis- tently marked by chromosomal alteration (CGL) not infrequently itself gives rise to fulminant acute leukemia with marrow blast cells arising by chromosomal alteration of the underlying Phl-positive clone, A final set of disorders also appears to predis- pose to leukemia, although not through mechanisms of overt chromosomal damage. These involve immunologic conditions; some o f them are hereditary in nature (ataxia telangiectasia and Bruton-type X-linked agammaglobulinemia), and others represent particular familial situations (immunologic abnormalities in rela- tives of patients with various forms of fympho- cytic leukemia or lymphoma) (Blattner et ai, 1976).These particular observations seem con- fined to malignancies arising from lymphocytic cell lines and therefore strongly suggest that inherited propensities for immunologic mal- function may involve predisposition to lym- phoid neoplasia as well. I t I I1 ,I I1 ,I I II' ;I IS( I Tt i E LEUKEMIAS -- 733 Radiation Radiation, at least o f the ionizing sort, has lotig tieen known to cause leukemia. The earliest evidence came from cases arising in radiation workers soon after the discovery of x-rays. It s&stquently has been shown that leukemia is particularly frequent among radiologists, altliough t h i s excess risk is now decreasing, presrrniCihly as safeguards in radiation praf-ticc iriiprove (Matanoski et al, 1975). C onsiderable epidemiologic research has fo( rrscd on radiation leukemogenesis. Irifvmation has conie from three types of human expow c ' nuc-lea reactions, medical therapy, and nicclic a l diagnosis. Whatever the source of information, three general features stand out. first, increased incidence wenis directly related t o r;ldiation dose. Whether this relationship is cvitirely linear, particularly at low dose levels, ,ind whether it involves any threshold dose twlow which leukemogenesis does not o c cur, rtvnaiti matters of some uncertainty. Second, L i t t v i t periods rariging from 2 or 3 years to 20 or t i i o w years can be expected between exposure and Irukeniia appearance. And third, mainly noti-lytiiphocytic varieties of leukemia seem to nrisc from radiation exposure. A consistent featttrc. in all sets of data is the absence of excess risk for CLL, although radiation exposure is associated with increased risk of ALL in young survivors of the atomic bomb. Nuclear Reactions C piderniologic follow-up of Japanesepopulations surviving the 1945 atomic bomb explo\ton\ in Hiroshima and Nagasaki have clearly sliown the leukemogenic potential of ionizing radiation (Beebe et al, 1978). All forms of Icwkcniia except CLL increased in incidence, tlic earliest cases appearing about two years Rfivr {lie hombing. Peak excess was reached aftw 5 to 10 years, and incidence at 30 years post-lmnb is s t i l l slightly greater than expected. Iatcnc y has appeared to vary with age at exposurf1 and with type of leukemia, shorter latenc ics Ijeing associated with CGL and younger ages. As person-years of follow-up accumulate in t t i r japanese experience, excess leukemia itic itlence has become discernible at total body ~ O S Plevels in the range of 10 to 20 rem. The tlosc-tesponse relationship appears linear for thr. most part, at least at higher dose levels. In recent years, data suggesting excess leukeniia incidence have also been published for otticr population groups exposed to products of nuclear reactions: workers in the nuclear industry, military personnel present at atmospheric nuclear tests, and populations in areas especially prone to fallout from nrrclear tests. In no instance are the data entirely convincing, ant1 further studies are in progress. In the c a w of persons employed in nuclear industries (Man- cuso et al, 1977; Anderson, 1978) anti in nuclear shipyards (Najarian and Colton, 1978; Evans et al, 19791, extensive dosimetry informa- tion makes it possible to make close coniparisons between exposure levels and ultiniate health outcome. For nuclear test exposures, although some dosimetry information i s available, records are incomplete, and questions exist concerning extent of internal radiation from ingested or inhaled radioactive material (Caldwell et ai, 1980). Similar uncertainties surround the levels of radiation received by populations in Utah, Nevada, and Arizona through fallout from atmospheric nuclear testing in the 1950s and 1960s. A recently suggested excess o f childhood leukemia in southwestern Utah, although not yet clearly established (Land, 19791, may well he compatible with human close-response data from other sources i f cumulative fallout doses ptove to be in the range of 5 to 10 reni (Lyon et ai, 1979). Medical Therapy Strong evid.ence for radiation leukemogenesis comes from epidemiologic follow-up of persons receiving therapeutic radiation for various medical indications. Many studies have been conducted in this area, the oldest and most extensive of which is the British study of men receiving x-ray therapy for ankylosing spondylitis (Court Brown and Doll, 1965). In this particular set of data, follow-up has been sufficiently long and cohort size has been sufficiently large to permit estimates of dose-response patterns, at least for doses over 100 rem. As in other studies, no excess incidence has been seen for CLL. Excess leukemia incidence has also been observed i n studies of patients receiving radiotherapy for thymic enlargement, menorrhagia (Smith, 1977), and polycythemia. Medical Diagnosis Evidence of increased leukemia incidence following diagnostic radiation exposures is contained largely in studies of childhood leukemia following prenatal x-rays. The increased sensitivity of rapidly dividing fetal cells i s felt to account for leukemia induction at fetal whole 734 - CANCER BY TISSUE OF ORIGIN body dose levels in the range of 5 rem. Recent analyses of a large continuing British casecontrol study of childhood cancer reconfirm t h i s relationship and indicate a distinct dosercsponse relationship based on numbers of x ray film exposures during pregnancy (Bithell and Stewart, 1975; Kneale and Stewart, 1976). I3ata from a similar American study suggesting that pre-conception radiation of parents may also increase risk of childhood cancer remains unconfirmed (Graham et al, 1966). Diagnostic radiation exposures after hirth have yet to tie firmly associated with excess lcukernia incidence. This is not surprising, of coursp, since cumulative dose levels are on the whole in the range of a few rads or less. Recent reniialysis of leukemia case-control data suggtlsting a substantial leukemogenic effect at thew low levels (Bross et al, 1979) has been srviously questioned on grounds of methodology (Boice and Land, 1979). The one exception among studies of postnatal diagnostic radiation may be the follow-up of persons receiving thorotrast (radioactive thorium dioxide), among whom c irmulative dose levels may well have Iwen suhstantial and who appear to show a tlcfinile increase in leuhemia occurrence (da Silva t lorta et at, 1974). Other Radiation Sources Several investigations have sought correlations between background radiation levels and hunian cancer occurrence. To date, no such correlations have been found, whether for leukwiiia or for cancers generally. However, since annual background dose levels rarely exceed 100 to 200 mreni, it is not surprising that no cancer effect has been seen. Based on doseresponse data from other sources, the linear response hypothesis would require investigation of exceedingly large population groups if cxc ess incidence were to be seen in populations in Colorado exposed to uranium mine tailings (Mason et al, 1972). A recent study suggesting a correlation between childhood cancer and magnetic field radiation in Colorado (Werth- eirrier and Leeper, 1979) has not been con- firmed in a similar study from Rhode Island (Fulton et al, 1980). Chemicals Various chemicals are known to be toxic for marrow cells, and it is not unlikely that many of these also possess leukemogenic potential. Evi- dence of leukemogenicity is strongest for benzene where recent occupational epidemiologic studies have suggested significant increases in leukemia incidence among workers with probable past exposure to benzene (McMichael et al, 1975; Infante et at, 1977). While not all such studies have shown increased leukemia incidence (Ott et al, 1978; Thorpe, 19741, the two studies suggesting an association between benzene and leukemia incidence support earlier anecdotal evidence from CASP~J of leukemia for which strong histories of exposure to benzene can be documented (Vigliani and Forni, 1976). While most data suggest that leukemias induced hy hen7ene exposure tend t o be myelocytic in cell type, one study concerning rubber workers showed an excess in lymphocytic cases (McMichael ef al, 1975). The latter study also suggested the possibility that organic solvents other than benzrne niight play a role in leukemogenesis. These observations, together with reports concrrning possible increased risk of leukemia in chemists (Olin, 1976) and possible increased marrow chromosome breakage in leukemia patients with histories suggesting occupational exposure to leukemogens (Mitelman et al, 1978; Lawler et al, 19791, point to a need for rnore epitleniiologic investigations concerning the leukemogenic potential of work-related chemicals. More indirect evidence of the potential leukemogenicity of organic hydrocarbons ( oines from a study showing excess leukemia mortallty among Nebraska farmers (Blair and Thomas, 1979), thereby suggesting a possible link to exposure l o chemicals used on farms. Data suggesting that leukemia and other cancers may be increased in children whose fathers work in hydrocarbon-related occupations (Fabia and Thuy, 1974) have not been confirmed in subsequent studies (Hakulinen et al, 1976; Zack et al, 1980). Chromosomal analyses, however, have raised the possibility of increased chromosome breakage both in chemical laboratory workers and in their children (Funes-Craviotoet al, 1977). Increasing evidence has accumulated in re- cent years concerning the leukemogenic potential of various therapeutic chemicals. Early an- ecdotal reports suggested such sequelae after treatment with phenylbutazone and chloram- phenicol. Later evidence focused strongly on the leukemogenic potential of alkylating agents such as melphalan used in cancer chemotherapy (Karchmer et al, 1974; Kyle et al, 1975; Coleman et al, 1977: Reimer et al, 1978; Berg- sagel et al, 1979) and busulphan (Stott et al, T t I E LEUKEMIAS - 735 .- 1977). Such reports underscote the need for servations. Pertinent evidence comes from a clinicians to consider the risk of treatriient- variety of sources, but in particular froin ( 1 ) induced cancer as they prescribe cancer thera- studies of leukemia occurrence in relation to py. (See Chapter 62 for further discussion.) known infectious agents and (2) studies of Despite the relative paucity of positive epide- possible clustering among leukemia ( asvs niiologic- findings with respec t to chemical Ieu- (or, in other words, thc. potenti;il relation o f kwiogcncsis in humans, i t remains likely, based leukemia to unknown infectious agents). on general knowledge from experimental an- imals, that such exposures do contribute sub- Infectious Agents stantially to case occurrence. Low-level expo- wres t o multiple chemicals, many of them with Numerous observations have been madp structural features similar to benzene, are fre- concerning the potential relationship of human quent in modern life. Specific exposures and leukemias to prior infection hy either human or specific dose levels, however, are extraordinari- animal infectious agents. Alniost without excep- ly difficult t o estalilish, and even when this may tion these studies have produced negative re- I)(.possihle, epidemiologic analyses are severe- sults, whether exposure has been hy direct ly hampered by limitations in sample size and contact with infected individuals or by inocula- by inability to correct for competing risk tion with potentially infectious niaterial such as fac tors In the absence of 'firm human data, vaccines. thcwfore, i t may be prudent to assunie that, as Particular attention has been given to expo- with exposure t o ionizing radiation, chemical sures in utero, under the hypothesis that tlw cui)osures rdate in a roughly linear non- immature immunologic status of the fetus niay thioclioltl nianner to leukemogenesis and t o make it particularly susceptihle to leukernogen- plan preventive strategies accordingly. ic infection and hence may account, by a latent period of two to three years, for the inciclcnce pcak that appears for childhood leukemia at Infection about age 3. initial observations in Gteat Britain suggested that in utero exposure to influenza It i s highly likely that infectious agents play a might have such an effect (Fedric k arid A l t w - major role in human leukemia etiology. The man, 1972). Subsequent studies, however, have viral etiology of nonhuman leukemias has been yielded conflicting findings, and it now sernis repeatedly demonstrated. Lacking the opportu- unlikely'that any consistent leukeniogenic elfcr t nity, however, to do the same kinds of exprri- exists (Hakulinen et al, 1973; Randolph and rncntal studies in man as are possil)le in rodents, Heath, 1974; MacKenzie and Houghtoii, 1974; c ,its, monkeys, and cattle, we are forced to use Curnen et al, 1974; Austin et al, 1975; Shore ct indirect means to seek evidence of infectious a!, 1976). Similar suggestions have Iwen tiiadp c.tiology in human leukemias. Thus far, these for varicella infection in pregnancy (Bithell ct al, efforts have yielded remarkably sparse results. 1973; Vianna and Polan, 19761, but obscrva- The tec hniques currently available for such tions thus far seem too few for firm conclu- studies, however, are limited to standard epide- sions. miologic approaches, successful enough for Considerable work has also been dcvotcct to c Inssicid infectious diseases in which latent the proposition that animal leukemia virusw periods are short, infectivity i s often high, and may be infectious for humans and t h u s may c linical illness is relatively conimon, but unlike- account at least partly for human leukemia ly t o be productive in studying infectious events occurrence. Animal tumor antihody stticlies in t h i t animal models predict will involve long veterinarians arid in other persons particularly lateticy, variable expression of infection, proba- exposed to animals with inalignaticy (primarily hlc vcrt ica I t ransmission from generation to cats and cows, but also rodents, primates, and gcnrr,ition, and relatively infrequent clinical other species) have been almost entirely nega- disease. tive (Heath et al, 1975). At present, therefore, In the absence of a clear understanding of there is little room to speculate that any signifi- human v i r 4 oncogenesis, and even with labo- cant animal-to-human tumor virus contagion ratory tools for detecting viral activity (e.g. exists, at least within the limits of current I a h - procedures for isolating and identifying in- ratory techniques. In a similar vein, the concept fectious agents or for measuring antibody re- that infectious disease immunizations might sponse levels), we must be content for the pres- lead to increased leukemia occurrence through ent with relatively crude epidemiologic ob- transmission of animal tumor virus contami- -- tr 736- CANCER B Y TISSUE OF ORIGIN nants (SV-40 virus in killed poliovirus vaccine hood leukemia cases (Caldwell and Heath, and avian leukosis virus in yellow fevw \:x-1976). Studies of interpersonal contact clustcr- cine) has not been supported by epidemiologic ing, while focusing primarily on Hodgkin's tiis- studies. Likewise, evidence concerning the pos- ease and related lymphomas, have yielded vari- sibility that BCG vaccination might protect able and as yet inconclusive results with respcct against leukemia development has heen largely to leukemias and to lymphomas. negative (Rosenthal et al, 1972; Snider et at, Other types o f case clustering have also t m n 1978; Skegg, 1978). sought without success. Leukemia occ iirrcw P Much speculation in the past has focused on appears to follow no consistent scvmnnl pat- infectious mononucleosis (IM), its cytologic terns. Likewise, no tendencies exist for Ivtrkc- similarities to leukemia, and the possibility that mias to occur with undue frequency in spoiises its etiologic agent, the Epstein-Barr (EB) herpes of leukemic persons, in children Ixmi to Iccr- virus, might also induce leukemia. Studies relat- kemic mothers, or in persons who receive hlood ing to this hypothesis have been consistently transfusions from individuaIs who Iater dcvt. I()p negative, although the situation appears dif- leukemia (Greenwald et al, 1976). Orie re( erit ferent for lymphomas; E6 virus i s likely to be a retrospective study suggesting inc reased frp- causative factor for Burkitt lymphoma, and quency of hospital-related exposure'\ among t todgkin's disease has been found to be signifi- patients with leukemia i s as yet unc onfirmcd cantly increased in persons with prior het- (Timonen and llvonen, 1978). erophii-positive IM (Rosdahl et al, 1974). Despite this mass o f generally nq"a, t 'IVC 01 The latter observation may well reflect altered inconclusive inforniation, one should not de- immune responses following IM rather than cide too quickly that inft.ction is not of ctiologic simple infection by EB virus. importance for leukemia. For many known in- fectious diseases, if statistical tests for clustcring Case Clustering alone were to be applied, it is unlikely that evidence would emerge t o suggest clustering or An extraordinary amount of epidemiologic infectivity. In the atisence of specific biologic effort has been devoted over the past 20 years to tests (molecular markers, virus isolation techthe possibility that leukemia cases, if not cancer niques, antibody measurements) arising from cases more generally, may tend to occur in knowledge of actual lcukemogenic me( han- clusters, thereby reflecting the undeilying isms, it seems unlikely that the crude cpidemio- operation of infectious agents as yet uniden- logic techniques currently available will 1 x 3 ahlc tified. While much that has been published to provide greater insights into leukemia causacon(-erning t h i s subject has consisted of anec- tion thaq are presently at hand. dotal accounts of individual case clusters, a substantial number of studies have sought sys- tematically to assess degrees o f clustering PROSPECTS FOR PREVENTION among cases in defined populations (Caldwell and Heath, 1976; Smith, 1978). While some As indicated in the introduction t o t h i s chap- work has continued to focus on cases clustered ter, immediate prospects are not good for pre- simultaneously by time and place of occurrence venting the occurrence o f leukemias t o any (Alperovitch et al, 1974; Smith et al, 1976; substantial degree. Despite extensive research Kraus et al, 1978), attention has tended to shift regarding pathogenesis and causative factors, more recently t o the concept of cases clustered our understanding of the origins of the leuke- hy virtue of past person-to-person contact, "ac- mias does not yet have the depth necessary for qiiaintance networks" with or without cases devising primary prevention measures. coinciding in time, in place, or jointly in time Two areas in which preventive programs may and place (Pike and Smith, 1974; Schimpff et al, be successful, however, involve curtailment of 1976; Zack et at, 1977; Greenwald et al, human exposures to man-made sources of ion- 1979). izing radiation and to various environmental In both instances (time-place and interper- chemicals with leukemogenic potential. In ei- sonaI contact cIustering), resuIts have suggested ther case, we would act in the tradition of John little if any tendency for cases to come in Snow whereby a preventive program does not clusters beyond what chance would predict. require precise knowledge of the etiology and Several studies, nonetheless, using a variety of pathogenesis of a disease. (The example i s statistical approaches, have suggested a weak overdrawn, of course, since Snow removed the degree of time-space clustering among child- offending pump handle after the cholera epi- T H E LEUKEMIAS - 737 . deniic was nearly spent.) For ionizing radiation, this piimarily nieans programs for controlling the medical and dental uses of radiation, since ~ C K11 exposures account for about 70 per cent of man-made and hence controllable radiation. For chemicals, prevention will need to focus on t)enzene, a common solvent and gasoline addi- tive, and perhaps other related compounds. ( hc u p a h i a l exposures will need particular attention, although the accumulated total population impact o f chemical exposures i s primarily a nonocc upationaI matter. Direct measurement of the success of SIJC h r.xi>osiir~~-contropirograms w i l l unfortunately n o t be possible. Since the leukemias, like can- c w 5 generally, are of multifactorial etiology, it i s impossil)le t o know which particular case results from which particular cause or set o f c mscs. Nonetheless, a decline in non-CLL leukwiiia5 at an appropriate interva: after instituting c ontiol of radiation and chemical exposures might bc presumptive evidence of success. In thc interval, achievement of documented re- d u c t i o n s i n exposure levels would need to srtffice as a nieasure of sciccess. Considerable prog~ess may already have been made, of come, at least with respect to ionizing radiation, through improvenient in radiologic techtiology and medical practice. Unfortunately, iinprovetnents in leukemia diagnosis a n d reporting at the same time that potential exposure to radiation may be occurring may have ob5( tired true declines in incidence. The variable Intcnt periods involved and the effects of treatment i n widening the gap between incidence and rnortality will also tend to confuse the issue. t i t the meantime, the ultimate possibility of w m c form of preventive strategy based on the rtnttcdying viral or immunologic nature of hutnan leukemias should remain a long-term goal. It seems unlikely that this possibility can I)c achieved, however, without sustained efforts in I m i c research concerning cell processes and on( ogenic mechanisms on a broad multi5pecieF basis. A fundamental recommendation, thwefore, for prevention of leukemias is to maintain strong support for etiologic research. SUMMARY 1I w leukemias account for a relatively small proportion of total cancer incidence but have rcccivcd a n exceptional amount of etiologic atid cpidemiologic research attention. Although specific cellular and molecular mechanisms of I~r~kemogenesriesmain unclear, the etiology of the leukemias can be characterized broadly in terms o f genetic factors, exposure t o ionizing radiation and chemicals such as benzene, and the potential influence of infectious agents. While long-range possibilities for priniary prevention of the leukemias depend greatly o n sustained basic research, particularly with re- spect to viral and host susceptibility factors, some immediate success seems possible through improved control over populat 'ion ex- posures to man-made ionizing radiation and to marrow-toxic chemicals. References A l p r o v i t c h A, Hesse C , Larar P, et al: Twnpmal-spati.il distribution of leukaemia and haematusarrnnia in w v e n Frenrh reginns. l r i t I Epidemiol 3:209-218, 1074. Anderson TW: Radiation exposure of lianfnrd worktys: A critique rif the hlancuso, Stewart and Knealr repnrt. Health Phyc 35:743750, 1978. Austin OF, Karp S, Dworsky R. et al: Excess leuhaniia in cohorfc of rhildren horn following influenza epidemics. Ani ) Epidrtnir~l 101:77-83, 1975. Beebe CW, Kato H, l a n d CE: Studies of the mortality of A-lmnih survivors. 6. Mortality and radiation dose, 1950-1074. Rndiat R1.c 75:138-201, 1978. Bergsagel DE, Bailey AI, Langley CR, et al: The thcrnotlierapy of plasma-cell niyelonia and the incidence of acute I w k r n r i a . N Fngl J M e d 301:743-748, 1979. Ritliell IF, Draper GI, tiorbark PO: Association between tnaliRnant disease in children and maternal virus infertiow. Rr Met1 1 1:706-708, 1973. Bithell IF, Stewari AM: Prenatal- irradiation and chiltlhrrod malignancy: A review of British data from the Oxfnrd survey. Br I Cancer 31:271-287, 1975. Blair A, Thomas TL: Leukemia amnng Nebraska farnicrc: A daatli rertificate study. A m J Epideniiol 110:264-274. 1979. Rlattner WA, Naiman IL, Marin DL. et al: IiiiniurioRc,li~,tir dt~ternii- nants o f familial acute lymphcwytir leukmiia. Ann Intern ht~tl 89: 173-176, 1978. Blattiier WA, Strober W, Murhninir AV. et aI: Faniilial i hrnnir lymphocytic 3ukemia - inrniunnlngic and c-cllul.ir rharac tcri7a. lion. Ann Intern M e d 84:554-557, 1976. Enice ID, Land CE: Adult leukemia fnllowing diaRntistic x-rays? (Review of a reporf b y Bross. Ball. and Falan on a t r i - m t e leukemia survey). A m I Public Health 69:137-145, 1979. Bross 101,Ball M, Falen 5: A dosage response curve for the otic rad range: Adult risks from diagnoslir radiation. A m I Puhlic t tealth 69: 130-136, 1979. Caldwell GG,lleath CW Ir: Case cluslerinR i n cancer. South M e d J 69:1598-1602, 1976. Caldwell G C , Kelley DB,Heath C W Ir: leukemia atnnng parliripants in military maneuvers at a nuclear bomb test (Snioky)., IAMA 244 :1575-1578, 1980. Chaganti RSK, Miller DR, Meyers PA, et al: Cytogenetic evidenre of the intrauterine origin of acute leukemia in monozygotic twins. N Engl I M e d 300:1032-1034. 1979. Colenian CN, Williams CJ, Flint A, rt '11: I i m i a l o l o g i r neoplasia in patients treated for Hodgkin's disease. N Fngl I Mtvl 297: 1249 - 1252, 1977. Court Brown WM, Chll R: M t m i l i t y from Cancer and o t h w rnusrc atter radiotherapy for ankylosing spondylitis. Br Med j 2:1327- 1332, 1965. Curnen MGM, Varma AOA, Chridinr- BW, et al: Childhnntl lroke- mia and maternal infectious diseases during pregnancy. l N C l 53:943-947, 1974. C:uh SI. Young JL jr: Third National Cancer Survey: Incitlrnce Dala. Natl Cancer Insf Monogr No. 41. 1975. da Silva Hofta J,da Mona LC, Tavares MH: Thorium dioxide effects in man. Environ Res 8:131-159, 1974. Draper GIvHeaf MM, Kinnear Wilson I M: Orcurrrnce of childhood I _. !e 738- CANCER BY TISSUE OF ORIGIN cancrrs aniring silis and rstimation of f,iniilial risks. Ihlrtl <;c.net 14:Rl -90, 1977. F\.inr I II, Buckton KE, 1 I.irriilton (;F, 1.1 dl: Kafliation-ititlucrd c-liroinomne aherrations in nu( bar-tlockyartl wrirkrrc N.iture 277:531-534, 1979. r.iliia 1, rlluy TI): O < r u p a t i o nof fathrr at tinie of I)irtli of r l i i l d r r n flyiiig of malignant diseares. Br I Prrv Sor M w l 28:98-100, 1`174. f.illrtta IM, Starling KA, Frrnliach 171: Lriikrnii.i iii twins. Pcdi.itric )I `.2:84h-D49. 1973. Frtlrirk I, A l h r n i m FI): Rcyxirtrd inf1urnz.i in prrgii.iw y ,m!! sirli.;rquf:nt crliirrr i n thr child. Hr hltd I 2:485--48R. 1972. riiltorr IP, C r h h S, Prd)Ir 1, rt al: Flc,c-trical wiring rwifigurdion\ .in0 ( hildhwid Iviikwii.i in Rhotlr Island. Ani 1 El)itLviiiol 1 1 1:F?2-2%, I9RU. riiiir~s-Cr,i\,ititf) F, Z.ipata-(;.iyon C, Kolniotlin-1lcdiiian R, rt <il: ( Iironiosonie alwirations *indsister-( hroniatid r x { h.ingr in workr r q in r h r r i i i c ,it Islwratriries and a lotoprinting f x t o r y t i ~ i c lin c Iiildrrn of wniiirn Ia1)oratt)ry workrrs. I .in( r t 2: 122-125, 1977. C;rah,irii S, 1.rviii M L , I iliviifrld AM, et ;if: Prrt om qHion, intr,iutt+ inr and pnctnat,il irradiation as relatrrl t o Iruktviii.i. Natl (-am PI Inst Moriogr 19:347-?71. 1966. Grrrnwaltl P, Rcirr IS, r).litr h PR: Ac qu.iint.inrc w t w m ks amwig Irukcmia ant1 lyniplxinra patirnts. Ani IFpitlrniiol 1 10:162- 177, 1`979. C;rwiiw.ild P.Woot1,irtl F, N,isra PC, rl al: hlorbitlily .ind nrorl.ility aniong reripicwts of hlood from preleukemic and pielynipho- matous donors. Cancrr ?R:.324-.128. 1976. IIakulinrn 1, t i o v i I ~ , i r k i i i ~ ~ ~ i - ~ ~ i ~ ~Ms ,k r~t, l`11,:i iA~sio~c~i~&i)n Iwtwren i i i f l t i i ~ i i 7 d~ u1 ring prrgn.inc y .iirtl r1iiltlh~n)tIlrwkaviii.i. Br hktl I4:265.-.?07,1973. Il a l i i l i w n T, S i l o i i r n I.Ir(ipo I.:('ant er in tliv o f f y i n g cif htlirr\ i n h ~ t l r c ~ . i r l i o t i - r c ~ ~ , iIl)l(l(f ~u~p~.ilions. Br I I'rcv Sot Mrtl 11): 1 38140, 1976. II r v t l i CW Ir: lhr rpitlrniiology n f Ieukcvnia. In Sc h o t t r n k ~ l r1l) (rrl): Can<vr Epitlrmiology and l'rc~vrntion. Springfic-Id, Illinois, (' C Thomas, 1975, pp. 3 1 R -350. Iiratli C W Ir: tirretlit.iry Lictrm i n I r u k r m i a antl lym~~liniirasIn Lynrh I i T led): Cancer (;c?rietics. Springfield. Illinois, Charles C Thomas, 1976, lip. 23?-248. Heath C W Ir, Caltlwrll GG, Feorino K: Virusm and otlicv n i i c ro1x.s. Iri Fraunimi IF Ir ffd):Persnns at t-ligh Risk of Canrrr. An Approach lo C.inrer Etiology ,ind Control. N r w York. Ac a d m i i c f'rrss, 1975. pp. 24 1-264. Irrfante PF, Rinsky RA, W.rgoncr IK, et al: Leukaemia in l r n z r n r wirkrrs. Lanrrt 2:76-78, 1977. Kart Iinier RK, Aniare M. Larscn WE, r t al: Alkylating agrnts as Iwikrniogens in mtiltiplis myrloma. Cancer 33: 1103-1 107, 1974. Kaur I, Catovsky D, Valcliniarscon H, et ai: Familial acutr myeloid I r u k m m i a with ac.quirrd Pelger-Huet anomaly antl aneuploidy n f (1 group. Br Mcd I4:127-.331, 1972. KnrI1Ie CW, Stewart AM: Mantel-tlarnszel analysic of Oxford data. 11. Indepentlenl effects of fetal irradiation whfactprs. lNCl 57: IIN)9-1014. 1976. Kraw IF, Franti CF, Melicharek M, et at: Childhtxxl Ieukrniia in a rrir.il county: I)nes "unirwal" i n r i d c w e signify clustering? Leuk R r s 2:97-103, 1978. Kirrita 5. Kamei Y: Genetic s of f m i l i a l Ieukrmia. ].ip 1 Iiurn Genet 14:163-1 79, 1969. Kyle RA, Pierre RV, Bayrd ED: Multiple niyrlonia ant1 acute Iru1Ptni.i a w x i a t c t l w i t h alkylating agents. Arch Int Mrd 115:185-142, 1975. Land CE: The halards of fallout or of epidcmiolr~giralresrarch Wdilorial). N Engl IMrd 300:431432, 1979. I;iwkr SD,Sumrnrrs;Kill EM, (:link tiMc 0,et al: Lettrr I n rclitcir. I mrrt 2:853--RS4. 1979. Lynn II.,Klauhtv MR, (;archer IW, et al: Childhowl leukemias associated with fallout from nuclear testing. N Engl I Mrd 3lM:397402, 1979. MacKenzir JS, Houghlon M: Influenza infections during prrgnanry: Aswciation with congenital malfornialions and with subwquenl nroplasnis in children and potential halards of live virus varcinw. Rart Rrv 3R:356-370. 1974. Manruso TF, Stewart A, Knealr C: Radiation exposures of tlanfortl workers dying from cancer and other causes. Health Phys 33:3693R5, 1977. h4asnn TJ, Fraumeni IF jr, McKay GW Ir: Uranium mill tailings and cancer mortality in Colorado. lNCI49:661-6h4, 1972