Document rwGjvLpDX7VB2d3a3zM0YZar

Medical and Pediatric Oncology 19:115-121 (1991) An Exploratory Study of Environmental and Medical Factors Potentially Related to Childhood Cancer JudithA. Schwartzbaurn, P ~ D ,Stephen L. George, P ~ DC, harles 6. Pratt, MD, and Bertha Davis, AS I To determinewhether a general-purpose maternal age, year of birth or diagnosis, f:e idemiplogicquestionnaire can be used in patient's race, and social class. The variables c ildhood cancer hospitals to identify asso- identified were length of time the patient ciations betweenenvironmental factors and was breast-fed !x2 = 16.1, P = 0.04); having the major types of childhood cancer, we a garden with fertilizers, herbicides, and repkt the results of the analysis of the data pesticides $ = 17.2. P = 0.03); maternal obtained from such a questionnaire. On use of sex hormones during the year before admission to St. Jude Children's Research the patient's birth (x2 = 18.2, P = 0.02); maHospital in Memphis, Tennessee. between ternal cigarette consumption (x' = 18.0. 1979 and 1986 a questionnaire was administered to 1,270mothers of patientsdia nosed P = 0.02): and patient contaa with persons with cancer (x' = 20.7, P = 0.01). Despite with childhood cancer. Approximatefv one- the large number of patients studied, we half of the children had acute leukemias identified fewer significant variables than ( n = 629); the remainder had lymphomas would be expected, on the average, under (n = 237) or s l i d tumors (n = W.,';;Q-t-h;e null hypothesis. We conclude that the sponses to questions regardingth'e p e t s ' . . data obtained from a general-purpose epi- and parents'enviionmentaland mediql$i,+ demiologic questionnaire do not provide a tories were compared across ninediagnostP usetul overview o i the association between categories. Onlt .5 of 232 variables remained exposure to environmental iactors and sev- nominallv statist.icallv significant iP < 0.05) eral tvpes of childhood cancer. after adjustingior contounding bv patient or Key words: questionnaire, epidemiology, risk factors INTRODUCTION and to identify the most important medical or environ- Environmental factors associated with the develop- mental variables. ment of one or more diagnostic categories of childhood -C include ionizing radiation [ 121, sex hormones MATERIALS A N D M E T H O D S [31. electromagnetic fields [4], nitroso compounds [SI, ad m a l prenatal marijuana consumption 161.Bemuse any particular diagnostic catesory of childhood cm rare.the large case-control studies necessary to 'dtahfy and evaluate these factors have been costly in money and time. We wished to determine whether Between 1979 and 1986, 1,929 new cases of histopathologically confirmed childhood cancer were diagnosed at St. Jude in the following categories: acute lymphocytic leukemia (ALL), acute nonlymphocytic leukemia (.ehJLL). Hodghn's disease. non-Hodgkin's a genrral-purpose epidemiologic questionnaire adminis- to the mothers of all newly diagnosed patients at a c u cancer hospital could yield useful etiologic From the Department of Biostatisticsand Epidemiology. Lioiversiry of -on. The advantage of this data collecuon is that it is more systematicthan a medical record Tennessee. Memphis. (J.A.S.I: Division of Biometp and Medical Informatics. Duke University Medical Center. Durinam. North Carolina 6 . L . G . J: Department of Hematology-Oncology (C.B.P.) and *W and less costly than a case-control study. We Division of Biostiltistics iB.D I . St. Jude Children's Research Hospi- to use the results to generate new hypotheses. tal. Memphis confirm previous observations. and identif) new enw- Received November 1.3. !990 -tal carcinogens whose effect was too small to be Address repnni requests Judith Schwartzbaum. PhG. Departmentof bm*cd clinically. The analysis reported below was Biostatistics and Epidemiolog! , Un:versit\ of Tennessee. Memphis. to summarize the salient features of the data sei T7\ L x l o . ~ m1O w i k y - ~ i s s .Inc. .L .I--. 116 Schwartzbaum et al. lymphoma (NHL), neuroblastoma, Wilms' tumor, os- T.ULE I. Topics In Questhumire teosarcoma. rhabdomyosarcoma, and Ewing's sarcoma. Patient's history Parent's history Tnese nine categories represent the childhood cancers having the highest national incidence [7]except for brain and central nervous system tumors (no brain tumor Medical Birth weight Congenital disorders program existed at St. Jude's at the time that data Breast-fed Medical Congenital disorders History of cancer Ulnesses * collection was initiated). All parents of patients with Tonsillectomy Hospitalization these diagnoses agreed to be interviewed. However, having a small staff precluded interviews of 659 parents, vielding a 65.8% overall participation rate. The age, sex, k d race of the children whose parents were interviewed Hospitalizations Radiation treatment .Medication Illnesses tnjuries Radiation Medication Reproductive events Contraceptive nlicit drug use ! did not differ significantly from those whose parents were Residential Occupation t not. Interviews were scheduled as w o n as possible after the child's diagnosis (median interval = 2.9 months) with nonsignificant differences in the intervals among the diagnostic categories. Data from admission interviews containing demographic information on patients and their families were combined with the patients' diagnoses. In addition, each patient's mother (or father in the small minority of cases when mothers were not available) was interviewed by a trained interviewer. Tne hour-long Birthplace Residence of longest duration Quality of drinking water Lived near factory Lived on farm Hobbies Fertilizer, herbicides. pesticides in garden and house Chemicals in house Occupation during and after the pregnancy Occupational exposure chemicals. dust, fumw and x-rays Tobacco and dlcohol consumption Parental and other household members I i e interview was based on a questionnaire that solicited information about the topics listed in Table I. The goal of the analysis was to jummarize the data set, and :he father's occupation (0 = blue collar workers, ! makmg as few ilnjusrifiabie assumptions as possible. We such as construction workers. assembly line workers, and compared exposure across the nine diagnostic categories truck drivers; 1 = technicians. clerks, and salesmen; I i nther than arbitrarily designating one diagnostic cate;org.j)s cases and the others as controls. If cancer cases m;;psed llsccontrois and both cases and controls are usqiated with a common exposure, the casexontrol comparison will show 110difference in the distribution of 2 = professionals, such as physicians, engineers, and scientists1. .A11 these potentially confounding variables :J were initially included in the regression equations a& .' ' then removed according to the change in estimate q&:''& rion [ I 11. No adjustment was made for multiple testing, i I I i exposure. This lack of difference may erroneously be because this is an exploratory analysis. Thus all P values intevreted as i10 exposure effect. are nominal P values whose interpretation is not the usual iTo select categorical variables from the 116created by one. the questionnaire, the Pearson chi-squared sta:istic was used to test the equality of each exposure variable among the nine diagnostic types of cancer. The categorical RESULTS variables identitied as differing among the diagnostic The patients whose parents were interviewed were i categories ( P< 0.03 were then stratified by patient or representative of the St. Jude population of newly maternal age (assubstantively appropriate)and tested for diagnosed cases between 1979 and 1986 with respect to i ! equality using the Cochran-*Mantel-Haenszeltest of sen-- age. sex. and race. Most of the patients were white, and e n l association [8]. Sixteen continuous variables based 428 were female (Table II). As would be expected from on the questionnaire were transformed to ranks. and the the national incidence patterns of childhood cancer (7, Kruskal-Wallis test [9] of the equality of the median patient sex. race. and especially age differences among I ranks over the nine diagnostic categories [IO] was the diagnostic categories were statistically significant. Iemployed. We then adjusted these continuous variables The variable selection procedures described in Mate- for age using an analysis of covariance. The adjusted rials and Methods identified tive statistically significant odds ratios for the remaining variables were estimated variables (Table III): 1) the number of months the patient using polytomous logistic regression. The odds ratios were adjusted for the patient's age at diagnosis (an index of recall bias for prenatal exposures). maternal age at time of the patient's birth (for prenatal exposure). or maternal age at the time of diagnosis (for childhood was breast-fed: 2) parental gardening with fertilizers, herbicides, and pesticides dunng the period from birth to diagnosis: 3) the maternal prenatal use of sex hormones; 4) maternal cigarette consumption during the index pregnancy; and 5 ) the patient's contact with persons with t I exposure). year of birth (or the year of diagnosis), the cancer during the period from birth to diagnosis. patient's race. number of years of maternal education. Rhabdomyosarcoma is used as the reference category Environment and Childhood Cancer TABLE 11. Charactetistio of 1,270 Childhood Cancer Patieats Whose Parents Were Interviewed Diagnostic category No. Age at interviewed diagnosisa White No. Percent Female No. Percent ALL ANLL Hodgkin's disease NHL Neuroblastoma Wilms' tumor Osteosarcoma Rhabdomyosarcoma Ewing's sarcoma Total 522 107 133 104 104 101 78 72 49 1.270 4.6 465 89.I 220 42.2 7.4 96 89.7 41 38.3 14.6 12 90.2 54 40.6 10.2 90 86.5 32 30.8 1.8 93 89.4 53 51.0 3.3 74 73.3 55 54.5 14.5 66 84.6 29 37.2 7.0 63 87.5 35 48.6 14.7 48 98.0 15 30.6 6.3 1.1 15 87.8 534 42. I 117 aMedian age in years. bKruskal-Wallistest 191of equality of ranks of ages. Test of equality of percentages. TABLE III. Odds Ratios* Adjusted for Confounding Variablest DirgnOStiC CpLCgory Rh.Momyosarcoma' ALL ANU. Hodgkin's disease NHL NeProbIrstoma Wilms' tumor Orteorucoma E*iag's sarcoma Tarl No. "vaoge Months breast-feda 1.O I .o 0.8 0.4 (.02P 1.1 1.4 0.9 0.6 2.2 357 4.0h ~ ' ( 8 1= 16.1' P = 0.042 :.-GWening ,~!,G-wjth -. ,Wsticides" I .o I .? 0.9 I .4 1.3 1.1 0.7 2.6 (.01)F 1.1 I2 53 43.79 X?8, = 17.2 P = 0.028 Prenatal Sex hormones' I .o I .? 0.5 0.5 0.8 0.8 1.1 0.7 I .5 1.237 12.56 &, = 18.2 P = 0.020 Maternal ci_earetie consumptiond 1.o 0.7 1.5 1.1 0.6 1.3 0.8 1.1 1.4 1.240 28.6% X2,8, = 18.2 P = 0.021 Patient contact cancer patients' 1.0 - -. 0.6 O.! L05P 0.9 0.9 0.7 0.5 (.04P' 0.4 c.O?)e 1.6 1.246 37.96 x:,8: = 20.7 P = 0.008 'btk of exposure odds for each diagnostic catego? to exposure odds for rhabdomyosarcoma. $-On polytomouslogistic regression using maternal age or age at diagnosis. year of birth or diagnosis. the patient's race. number of years aDuaarl education, and the father's occupation as potentially c o n f o d i g variables. h a v e r a g e exposure is defined as being breast-fed for 8 monthsor more(the third quartilei:below average exposure IS defined as being k - f dfor2 months or less (the first quanile). 'ugardening wich fertilizers, herbicides. and pesticides from birth to diagnosis. w of KXhormones from 1 year before the pregnancy to the time of birth. pf "1" if ever smoked cigarettes during the pregnancy and " 0 otherwise. during the period from birth to diagnosis of the patient's index cancer. ip*bcategory. Of less than 0.05 for chi-squared test of equality of odds ratio to 1.O. number of months breast-fed. Overall association between exposure and diagnostic categor). (* M O[OR]= 1.O) for the eight other diagnoses. length of time a patieni was breast-fed was reported ad . a Y z e d in months. The odds ratios for breast- h g Shown 111 Table 3 compare the exposure odds of b - f d i n g for 8 months (the 75th percentile, relative e b t - f e e d i n g for 2 months (the 25th percentile) for each of the eight diagnostic categories separately with the exposure odds for rhabdomyosarcoma. The remaining variablesin Table 3 are dichotomous (coded 1 = yes and 0 = no) Maternal cigarene consumption is considered positive if the mother reported smoking cigarettes at all during the prenatal period. . '* :9 .. ri 118 Schwartzbaum et al. The five statistically significant variables that we overall P value of 0.02. none of the nine categories identified are discussed in order of their epidemiologic individually had a P value less than 0.05 (for testing an plausibility based on published research. Overall. few of odds ratio of 1.0). The age-adjusted proportions of the odds ratios diffec greatly from 1,0, and only five of mothers who smoked at least 10 cigarettes a day com- the odds ratios are associated with P values of less than pared with those who smoked fewer than 10 cigarettes a 0.5 (Table III). day and with those who did not smoke showed little Number of Months Breast-fed vahation among the diagnostic categories. Stjernfeldt et al. [13] found an association between Whether the children had been breast-fed was not maternal prenatal cigarette consumption and ALL but M) related to the diagnostic categories. However, the dura- relationship between maternal cigarette smoking and tion of breast-feeding differed among the groups. We childhood solid tumors. To replicate their analysis, observed a statistically significant protective effect patients with solid tumors (neuroblastoma. Wilms' tu- against Hodglun's disease (OR = 0.4, P = 0.02) m o n g mor. osteosarcoma, rhabdomyosarcoma, Ewing's sar- children who were breast-fed at least 8 months compared coma) were used as a control goup for ALL patients, and with children who were breast-fed no more than 2 confounding variables were controlled by unconditional months. Breast-feeding children at least 8 months did not binomial logistic regression. Our analysis produced odds offer protection against NHL (OR = 1.1, P = 0.73). ratios of 0.8 (P= 0.04) for smokers of 1-9 cigarettes a Because breast milk may be a vehicle for carcino_gens day compared with those who did not smoke and of 1.1 [12.131, we assessed the modifying effect of breast- C P= 0.48) for those who smoked 210 cigarettes a day feeding on maternal cigarette consumption and use of compared with those who did not smoke. fertilizers. herbicides. or pesticides. We found no evi- dence for a relationship between breast-feeding and these Patient Contact With Persons With Cancer During two variables. the Postnatal, Prediagnostic Period Parent Gardening With Fertilizers, Herbicides, and Pesticides During the Postnatal, Prediagnostic Period The highest odds ratios for patients who had previous contact with persons diagnosed with cancer were found amon9 patients with Ewing's sarcoma iOR = 1.6, . The !owest odds ratios for pudening with fertilizers. P = 0.281and rhabdomyosarcoma (OR = 1.O. reference herbicides, and pesticidesafter the birth of the child until disease). Approximately 4% 1313) of the identified the date of diagnosisswere reported by the parents of contacts reportedly had the most common types of adult children with Wilms'-tumor (,OR= 0 . 7 , P = 0.30, per- cmcer: gastrointestinal. lung, and breast. Almost one- centage exposed = .3(%.8)p:arents of children with os- half of these contacts were the patients' natural _ m d - teosarcoma reported the highest exposure (OR = 2.6. parents. Because previous investipors [ I ]have identi- P = 0.01. percentage exposed = 33.6,. Rdated vari- fied clusters of childhood leukemia and lymphoma, the ables that were not statistically significant included proportion of reported leukemia or lymphoma contacts of whether pesticides were used around the house. whether the study subjects with these diagnoses were compared the family !ived on or near a farm where crops were with that among subjects in the other diagnostic catego- sprayed. and whether either one or both of :he parents ries. No differenceswere seen. with approximately 5% of were farmers. the index cases of both groups reporting contact with Maternal Prenatal Sex Hormone U s e leukemia or lymphoma patients. Of 155 mothers who reported using hormones during the year before the index birth. 8 used fertility agents. 132 used birth control pills, 26 used progesterone. and 19 used a hormone they could not identify (some women used more than one type of hormone). The odds ratios range from 0.5 ( P = 0.1 1) for mothers of children with Hod_gkin's disease to 1.5 ( P= 0.55) for mothers of children with Ewing's sarcoma. Maternal Cigarette Consumption During the Index Pregnancy The three largest odds ratios were for mothers whose children were diagnosed with ANLL (OR = 1.5. P = 0.20). neuroblastoma (OR = 1.3. P = 0.40). and Eaing's sarcoma (OR = 1.4, P = 0.39). Despite an DISCUSSION In this exploratory analysis it is noteworthy that, after adjustment for confounding, few variables were found to be statistically significant. At the nominal 0.05 level, we would expect. on the average, approximately 12 (0.05 x 232) statistically significant variables to emerge by chance. We observed only five statistically significant variables and did not adjust for multiple testing; one interpretation is that the overall results are consistent with a global null hypothesis. This interpretation is Wen,@- sned by the fact that most of the odds ratios are near the null value (Table 111). The small number of statistically significant variables after adjustment for confounding may suggest the operation of common etiologic agents Environment and Childhood Cancer 119 among the nine diagnostic categories of childhood can- solicit information on pesticide use, only gardening with cer. Another cause could be time of exposure during the fertilizers. herbicides, or pesticides was statistically sig- prenatal period or childhood. If time of exposure relative nificant. Convincing evidence exists in the literature for to the development of the fetus or child is critical in an association between several types of childhood cancer prenatal and postnatal carcinogenesis [15.161, then con- and pesticide use. Buckley et al. [22] reported an founding by time could effectively remove evidence of association between prenatal and postnatal household and the influence of risk factors on tumor initiation or garden exposure to pesticides and childhood ANLL. promotion. Independent of this association. they also observed a Although we identified fewer variables than would relationship between parental occupational exposure to have been expected on the basis of chance alone. any one pesticides and childhood PLNLL. Lowengart et al. [21] of the five variables (or other variables) may be a true found parental use of pesticides in either the home or determinant of one or more of the nine diagnostic garden during the pregnancy and nursing period to be a categories of childhood cancer. Because there is no predictor of childhood leukemia case status. Pesticide statistical method for identifying which of a group of exposure has also been reponed in association with variables are statistically significant because of alpha neuroblastoma [23]. Wilms' tumor [24], and childhood error (false positives) and which represent a true associ- brain tumors [ 2 5 ] . In our study the odds ratios for the don, the results must be compared with those from children diagnosed with leukemia (ALL [OR = 1.3. previous studies to evaluate them. Of course, this method P = 0.383. AVL [OR= 0.9. P = 0.671). Wilms' tumor d l not identify previously unreported associations. (OR = 0.7. P = 0.30). and neuroblastoma (OR = 1.1. Davis et al. [ 171 observed elevated odds ratios for all P = 0.78, are not particularly high. However. that 'Ypes of childhood cancer for mothers who fed their associated with osteosarcoma is elevated (OR = 2.6, children nonhuman milk relative to mothers who breay - P = 0.01 1. The differences between our results and those fcd their children for at least 6 months. This odds r q " , f o u n d in the literature may be explained. in part. by the Was particularly pronounced for childhood Ivmph&a,,association of pesticide use with more than one site of and was probablidue to an association with chiidhiid "childhood cancer. Such an association could also account Hodgkm's disease [ 181. The association between the for the lack of statisticai significance of the other. duration of breast-feeding and ALL. SHL. and ANLL variables designed to evaluate information on pesticide Was not observed by Magnani et al. [ 181. and it was not use. Although Gold et al. i25] observed an association coafirmed for ALL by Van Duijn et al. I191. McKinne! between childhood brain tumors and household insect et d. [20] reported no difference betweer! the frequency extermination when the- compared the brain tumorcases with which the mothers of childhood leukemia and with nondiseased controls. they found no evidence for l-orna cases and their controls reponed breasr- this association when comparing brain rumor cases with facding their children. However. they did not repon the childhood cancer controls. wts of an analysis of the association between the The evidence for an association between maternal chuation of breast-feedingand leukemia case status. Like hormone consumption and childhood cancer. other than & b y et al. [20]. we found no association between adenocarcinoma of the vagina 131. is inconsistent. Three daldhood cancer and whether a mother said she breast- relativel) large. well-defined case-control studies. one of ber chdd. However. our results are consistent with acute leukemia [26], one of neuroblastoma 1271, and one b e o f Davis et al. [171, who showed a protective effect of Wilms' tumor [24]. found evidence for an association dbrrast-feeding against Hodghn's disease. and with the between maternal hormone status or consumption and wts of others I18.191 who observed no effect of childhood cancer. In the present srudy. we found a b - f e e d i n g on the childhood leukemias and NHL relatively low odds ratio (0.8)associated with the moth- [18*191O.ur findings are probably not attributable to ers of children with neuroblastoma who reponed prenatal bias, because every mother interviewed had a child hormone use. Robison et al. ! 6 ] .Operskalski et al. (281. with cancer. It has been hypothesized that and McKinney et al. (201found no association between b n olik consumption during infancy ma! increase maternal prenatal hormone consumption and ANLL. to early infection. which could indirectly offer osteosarcoma. and leukemia or lymphoma combined. -On against the subsequent development of respectively. n e overall percentage of women in this Halghn's disease 171. In addition. the effect ofbreast- stud) reponing use of sex hormones during the year %g may vary among populations dependin,0 on the before pregnancy (8.2%t was approximately the same *ntally determined constituents of breast milk ( 5 % t as that observed in a similar stud!. of childhood population. In some populations breast-feeding cancer cases conducted by Li et al. 1291. Further studies a vehicle for postnatal exposure to carcinogens that examine the time of sex hormone consumption (Q1321]. during the pregnancy shouid be conducted. Ot& several variables based or!questions designed I C Because no dose-response relationship uas observed ;.: r, 8- ' -4 I. -5 3 - --<1 ., ->* -3 .3 .t 3 ;: i i-q, -9 ,- I 120 Schwartzbaum et ai. for maternal cigarette consimption. the evidence for this REFERENCES variable in this study is weak: however, a case-conuol study identified a significant association between maternal prenatal cigarerte consumption and childhood cancer at all sites combined and ALL [ 14). Although two large 1. Stewart A. Webb J, Giles D, et al.: Malignant discw childhood and diagnostic irradiation in utero. Lancet 2447, 19%. 1-. Greenberg RS. ShusterJL Jr: Epidemiology of cancer in children. Epidemiol Rev 7:27-48. 1985. childhood cancer studies found no evidence for a similar 3. Herbst XL. Ulfelder H. Poskanzer Dc:Adenocucinoma of the association [30,3I]. further investigation would be worthwhile because of the re!ativeIy high prevalence of cigarette consumption combined with its known ability IO harm the developing fetus [32] and its carcinogenicity in vagina: Association of maternal stilbestrol therapy with apparance in young females. N En$ J Wed 18J.878-881, 191. 1. Wertheimer N. Leeper E: Electricd wiring configurations childhood cancer. Am J Epidemiol 109:273-284. 1979. 5. Preston-Manin S: N-nitroso compounds as a cause of h u m adults. The three statistically significant odds ratios for pa- tient contact with persons with cancer (Table 3) may result from the relativefy high frequency with which cancer patient contact w;1s reported by rhabdomyosar- cancer. IARC Sci Pub1 34:477484. 1987. 6. Robison LL. Buckley JD, Daigle ZZE. Wells R. Benjamin D, .Arthur DC. Hammond GD: Maternal drug use and risk of childhood nonlymphoblastic leukemia mong offspring. C e 63: 1904-191 I . 1989. 7 Young JL. Ries LG. Silverberg E, Horn Iw,Miller RW. C- coma c s e s (the reference category). Li and Fraumeni incidence, survival. and mortality for children younger than [33] and others [ 3 - 3 6 ] observed an excess of malignant neoplasms in the families of children diagnosed with both bone and soft tissue sarcomas. Although the odds ratio for wing's sarcoma is greater than 1.0 and consistent 15 yem. Cancer 385984502. 1986. 8 Lmdis Ill, Heyman ER. Koch GC:Average panial associationin three way contingency tables: $1 review and discussion of alternative iests. Lnt Stat Rev 46:137-354. 1978. 9 Conover WJ: "Practical Yonparametric Statistss." 2 4. New with these observations tOR = 1.6. P = 0 . 3 ) . that for York: John Wile? & Sons, 1980. osteosarcoma is less than 1.0 (OR = 0.4. P = 0.02). IO. Kirk RE: "Experimental Design: Procedures for the Behavioral Science," 1cd. Bclmont. CA: BrookCole Publishing Company, 1981. 1 I . Greenland S: Modeling md .ariable selection in epidemiologic CONCLUSIONS .- We conclude hat of the five variables identified. further research.&. the asscxiation between duration of breast-feeding and development of childhood Hodgkin's analysis. .bJ Public Health 793.10-349. 1989. 13. [nternational .Agency for Research on Cmcer: Tnchlorwthylene, tetra-ihlomethylene. alrbon tetnchlonde. LiRC Monogr Eval Cminog Risk Chem Hum 30:491-j14. 1979. 13 Barnell PC.EIIenberger HA: Obstructive ,aundice due to a chlonnated hydrocarbon in breast milk. Can Med Assoc J diseuse is most justified. However. despite [he large 117: I047-IO48. 1977. number of patients studied. we have idenritied few strong 1-4. Stjemfeldt M. Lindsten J. Berglund K. Ludvigsson J: M a t e d leads for further research. For conducting exploratory studies of several ppes of childhood cancer sirnulta- neously the general-purpose epidemiologicquestionnaire taken ;1s a whole is unlikely to yield sufficient benefits for smoking juring pregnancy and risk childhood cancer. k t I W 4 1: 1350-I252. 1986. i5 Yapalkov YP: Some general considerations on .he problem of rransplacenmi arcrnogenesis. In Tomarz L. bluhr U. Davis W &I: "Transplacental C;ucino_eenesis."Lyon: IARC. pp. 1-13. the: cost and effort involved. Further malyses of these data will determine whether in depth studies of particular diagnostic categories will provide usehl information about the epidemiology of childhood cancer. I071. 16. Movlgavkv SH. Knudson AG: Muration md cmcer: .A model for human carcinogenesis. J Natl Cancer Illst 66:1037-1052. 1981. I ? . Davis MK.Saviu DA. Gnubard BI: Infant Feedinn andchildhood ;mcer. Lancet 2:365-368. !988. IS. Magnmi C. Pdtore G. Temcini B: Infant feeding and childhood cancer. Lancet 2: 1136. 1988. ACKNOWLEDGMENTS 19. Van Duijn CW. Van Steensel-Moll HA. Vander Der Does-Vd. Berg A. Van Wering ER. Van Zanen GE. et d.;Infant feeding ad This research was supported. in part. by an Americ3n childhood cancer. Lancet 2:?96-797. 1988. 20. McKinney P.4.C m r i g h t RA. Saiu JMT. Wann JR. Stiller CA. Cancer Society Institutional Grant distributed by the Dnper GJ, Hanley AL. Hopton PA. Birch JM. Waterhouse JAH. University of Tennessee. Memphis. Cancer Center; Can- Johnson HE: The inter-regional epidemiological study of C h i l d cer Center Support (CORE) grant CA2 1765: Childhood Solid Tumor Program Project grant C.43099;and Lsukemia P r o p m Project srant C.420 180from the National Cancer Institute and the American Lebanese Syrian hood cancer (IRESCC): .A case control study of aetiolof$d factors in leukemia snd lymphoma. .4rch Dis Child 62:279-287. 1987. 2 1 . Lowengm RX. Peters JM. Cicioni C. et al.: Childhood leukernit and parents' occupational and home exposures. JNCI 79:3pJ6- Associated Charities (,%SAC). The authors thank Diane Cunningham for assistance in prepannz the manuscnpt. Barbara Kuyper. PhD. for editorial assistance. Richard Dodge for computer programming. and David Facklam and Lisa Fields for designing the questionnaire. 1987. -1-7. Buckley JD. Robison LL. Swotinsky R. Gambrant DH. M.Manchester P. Neshit ME. Odom L. Peters JM, Woods WG: Occupational cxposures of parents of children with acute nOntylp. phocytic leukemia: A report from the children's cancer swfv group. Cancer Res 19:403O-u)37. 1989. I Environment and Childhood Cancer 121 23. M n t e PF, Epstcin SS. Newton WA Jr: Blood dysrasias and childhood tumors and exposure to chlordane and heptachlor. scpad J Worlc Environ H d t h 4:137-150. 1978. 24. olshpn AF: "A Cuc-conbol Study of Risk Factors for Wilm's Tumor."University of Washington. DAI[Bl 49(4~:1I1S. 1988. 25. Gold E, Gordis L. Tonascia 1. Szklo M: Risk factors for brain tumors in childnn. Am J Epidemiol 1W309-319. 1979. 26. Van Steenxl-MollHA. Valkenbq HA. Vandenbroucke JP. Van Zantn GE: .etmaternal fertility problems related to childhood leukemia? Int J Epidemio! 14555-559. 1985. 27. KramerS, Ward E. Meadows AT. Malone KE: Medical and dmg risk factors associated with neuroblastoma: A case-control study. m a 78:797-804. 1987. 28. Operskalsh EA, Preston-Manin S. Henderson BE. Visscher BR: A uscconml study of osteosarcoma in young persons. Am J Epldemiol 126:118-126, 1987. 29. Li FP. Jamison DS. Meadows AT: Questionnaire study of cancer &lo@ in 503 children. JNCI 7631-36. 1986. 30. McKinmy PA. StillerC: Maternal smoking during pregnancy and tbe risk of childhood cancer. Lancet 2519. 1986. 31. Buckley JD. Hobbie WL. Ruccione K. Sather HN. Woods WG. Hammond GD: Letter. Lancet 2519-520. 1986. 32. Kline J. Stein Z: Circumsrances of exposure and reproductive consequences. In Gordis L red): "Epidemiologyand Health Risk Assessment." New York: Oxford University Press. pp. 25&267. 1988. 33. Li FP.Fraumeni JF J r Soft-tissue sarcomas. breast cancer and other neoplasms: A familial syndrome. Ann Intern Med ?1:747752. 1969. 34. Blamer W.4. et al.: Genealogy of a cancer family. J Am Med Assoc 241:25%261. 1979. 35. Birch JM. Hartley AL. Marsden HB. Harris M. Swindell R: Excess risk of breast cancer in the mothers of children with soft tissue sarcomas. Br J Cancer 49:335-397. 1984. 36. Hanley AL. Birch JM. McKinney PA. Teare MD. Blair V. Carrene J. Mann JR. Draper GJ. Stiller CA. Johnston HE. Ciutwri_ehtRA. Waterhouse JAH: The inter-regional epidemio- logical stud? of childhood cancer (IRESCC):Case control study of children with bone and soft tissue sarcomas. Br J Cancer 58:838-842. 1988