Tobacco Documents Online

REPORTS ) Environmental Tobacco Smoke and Lung Cancer Risk in Nonsmoking Women Hc:atllcr G. Stocktvcll, * AIlrnt L. Goldman, Gary H. Lyntctn, Charles 1. Nuss; Adum W. Armstrong, Patricia A. Pinkham, Elizabeth C. Cctndclnru, Marcia R. Brusa Background: Exposure to environ- mental, tobacco smoke (passive smok- ing) has been suggested to be a cause or lung cancer, although early epide- miologic studies have produced in- consistent resul!ts. Purpose: We conducted an epidemiologi'c case- eontrol study to assess the relation- ship between exposure to environ- mental tobacco smoke and lung eattrcer risk among women who have never smoked (i.e., having smoked for a total of <6 months or having smoked <100 cigarettes in their life- times). Methods: Case patients (n = 210) were women with histologically eon6rnted primary carcinomas of the lung who were lifetime nonsmokers. They were identified through hospital tumor registries and the Florida Can- cer Data System of the Statewide Cancer Registry: Community-based control women (n = 301) were allso lifetime nonsmokers and were identi- fied through random-digit dialing. Details on childhood and adulthood exposures to environmental tobacco smoke were ascertained through interviews with the study participants themselves or with surrogate re- spondents. Risks were calculatedin terms of smoke-years, defined as the sum of the reported years of exposure to cigarette smoke from each smoker Vol. 84l No. 11;J September 16. 1992 ' in the household. Resultr- The risk of lung cancer more than doubled for women who reported 40 or more smoke-years of household exposure during adulthood (odds ratio [(1R) = 2.4; 95%confid'ence interval [Cl[ = 1.1-53) or 22 or more smoke-years of exposure during childhood and ado- lescence (OR = 2.4; 95% Cl = 1.1-5.4). Risk.s were highest for non- adenocarcinoma lung cancers, al- though modest elevations in risk were also observed for adenocarcinomas. When: a surrogate respondent other than the patient's husband provided information on exposure, the risk esti- mates were considerably lower. Con- ctusion: These findings suggest that long-term exposure to environmental tobacco smoke increases the risk of lung cancer in women who have never smoked. [J Natl Cancer Inst 84:1417- 1422, 19921 In 1986, reports by the Surgeon Gen- eral (1) and the National Academy of Sciences (2) concl!uded that involuntary smoking can cause lung cancer in non- smokers. More recently, the National Institute for Occupational Safety and Health released a bullelin that reached similar conclusions (3). Several addi- tional studies have examined the rela- tionship between environmental tobacco smoke and the development of lung can- cer in nonsmokers. In 1991, Fontham et al: (4) reported the results of the first 3 years of a large multicentered' study of lung cancer risk among female lifetime nonsmokers. For nonsmoking women living with a spouse who smoked, there was a 50% increase in risk for develop- ment; of lung adenocarcinomas. In addi- tiont there was a significant increase inn risk for nonsmoking women exposed to: environmental tobacco smoke at work or during social activities. No associd- tion was noted for nonsmoking women ® exposed to environmental tobacco smoke during childhood. In contrast, Jancrich et al. (5) found that exposure to high, levels of household smoke during childhood and adolesccnce doublc& thc risk of lung cancer among male and female nonsmokers, whereas exposure to smokc during adulthood was not as- s>ciatcd with an excess risk. Othcr au- thor:. (6-8) notcd' an increase in lung cancer risk among tihc offspring of smokers, but their study populations in- cludcd few children who were not ciga- rettc smokers themselves. The current, study further explores the effect of environmental tobacco smoke and other risk factors for lung cancer on women who have never smoked to- bacco. Patients and Methods A population-based case-control study was conducted in central Florida to examine the causes of lung cancer in nonsmoking women. Women were elig- ible for inelusion as case patients if they had a histol'ogicallly confirmed primary Received February 24, 1992: revised May 2, 1992; accepted June S, 1992. Supported by Public Health Service grant CA-45513-U5 from the National Cancer Institute. Natiunal tnstitutcs of Hcaltht Department of Health and Human Services.. H. G. Stuckwell, A. W. Armstrong. P. A. Pinkham, E- C. Candelura, M. R. Btusa (Depart- ment of Epidimiology and Biosutisties), A. L. Goldman. G. H. Lyman,(Department of Internal Medicine, College of Medicine). C. I. Nos.a (De- partment of Environmental and Oeeupational Health. College of Public Health). University of South Florida. Tampa„ Fla. We thank all of the hospitals in the area who participatcdin the study and theStatewideCanecri Rucistry• Departmentof Health and Rehabilitative Services, Tallahassee„ Fla. The authors assume full, resprmsibility for thcanalvxs and interprete- tiun of the data. 'Cor.rrspnndcncc to: Headhct G. Stockwcdl- Sc.D.. Department ofiEpidrmiulog% ' and'Biosutis- tics„Collcge of Public Health. Unittasity of South Florida. Tampa, FL 33612-3R05. REPORTS 1417 'atwAin~'

carcinoma of the lung (internatiortal Classification of Diseases for Oncology codes 162 ?-162.9) that was diagnosed between April 1, 1987„and February 28, 1991. and if' they resided at the (ime of diagnosis in a 2tS:county area in central Florida. These women were identified Ihrouch the tumor rrgistrias of arca hos- pitals and the Florida Cancev Data Sys- tem of the Statewide Cancer Registry. Control subjects were community based and were idc'ntified through random- digit dialinr. AII case patients and con- trol subjects were lifetime nonsmokers. defined as having smoked for a total of less than tr month% or having smokcd' lc:a th•an Illt0 cigarettes in their lifr- time.. The smoking status of potential case patients in the study was confirmed at several aagcs. Once these individualss were identified by their hospital or the Statc%tide ('ancer Registry, rtcords„thcir smoking status was confirmed when their physician was contacted for per- mission to interview, again at the time of initial contact with the patient or next of kin. and; finally, at the eommence- ment of the intervieu% In addition, the interview contained questions regarding experimentation with tobacco, designed to elicit in a neutral manner any prior undisclosed tobacco use. Any potential case patient whose smoking status could not be confirmed was excluded. Of those found to be eligible. 83% of the ease patients or their next of kin agreed to be interviewed. The smoking status of control subjects was determined dur- ing random-digit dialing and verified during the interview. Trained interviewers interviewed case patients an& control subjects either in person or over the telephone. When nec- essary: questionnaires were mailed. Of the ease patient interviews, 41% were obtained by iin-per}on contact, 51% by telephone, and 8% by mail. Of the con- trol subject interviews. 518% X were ob- tained by in-persorr contact, 45.9% by telephone, and Q:3°k by mail! Informed consenn.vas obtained; from the case pa- tients and control subjects prior to the interviews according to the guidelines of the University of South Florida Ilnsti- tutionalL Review Board., The interview included questions on environmental, to- bacco smoke exposures at home, on thc job,, and in social scttings. If case pa- tients were too ill to be interviewed or were deceased, surrogate respondents (primarily husbands and ehildren) were interviewed. Interviews of surrogate re- spondents were necessary for 66.7% of the case patients. Odds ratios (ORs) were calculated to estimate the relative risks. Multiple to- gistic regression analyses were per- formcd using the SAS LOGISTIC pro- cedurc (SAS Institute, Cary. N.C.). Ninety-five percent confidence intervals (CIs) were calculated from the logistic models. The chi-square statistic was used to test for trend. Results The study population consistc&of 2111 women with histologically confirmed primary lung eanecr, who had never smoked (case patients) and 301 community-haud cont'roli women who had never smoked (control subjects). The basic demographic characteristics of the case patients and control subjects are presented' in Table 1. Ninety-three percent of the case patients and 94% of the control subjects were White. (The small percentage of non-White studyy participants reflects the small percent- age `596) of non-White case patients identified by hospitals and the statewide cancer registry.) Case patients tendcd& to be somewhat older and had fewer years of formal education than controls, with a significant trend of increasing lung cancer risk with decreasing years of schooling (P =.(119). Almost half of the case patients and control subjects had lived in Florida for at lcast 2(1 years. Table 2 indicates the ORs and 95% . Cls associated with exposure to ctga- rettc smoke from parenls, siblings, hus- bands, and other household members, after adjustmcnt for age. race, and education. Unexptnc& individuals were those with no household environmental tobacco smoke exposure. We doticrihce exposure in terms of smoke-year., dc- fined as the sum of the reportcd years of exposure to cigarette smoke from each smoker in the household. We considered smokc-ycars to he a more rcliahte meas- ure of exposure t'han pack-ycars, since study'participants had less difficulty re- calling the number of years they had lived with someone who smoked thann recalling the number of cigarettes per day to which they had been exposed. Smoke-years were subdivided into three categories of approximately equal size Tabk 1. Distributinn of selceted eharaneristic's of case patients and control subjcct. Gse patients (n . 2r0)' Control subject+,(n - V0I)• Characteristic No. % No. CA Race Whitc 196 93 21Ft 94 Non-White 14 , 7 117 6 Binh year Rcfuru 19211 t}7 65 179 MI I9Nt-1424 5: 2:5 7:1 -'a 14i11 or latcn 21 :' a h 49. 16 Marital status Married I'n>^ 51 174 58 other 1'0_ 49 127 42 Ycars of education cR gradcs U ' Ili 37 12 9-1 1 Ende.i 16 17 401 13 a 12 grades 1;5 65 22? 74 Year in Florida <Itu 47 22 65 22 10-14 63 V1 89 30 2u-29 .1-1 16 46 15 sil1 b6 31 101 34. Lung pncer cell types Adenoeareinoma 129 61 Squamous , eell' carcinoma ;6 17 Small-eell carcinoma 14 7 Alllother 31 15 ~ - - •Valucs may not add to total becaucc of missing data i Journal of thc Matiiroal Cancer In•titw. 1418

for both early and adult years. The dis- tribution of smoke-years of exposure was much lower for early years, pri- marily because participants tended to live with spouses for more years than they had lived with their parents, result- ing in lower culoff points. Lung cancer risk estimates for women who were exposed to environmental tobacco smoke during childhood and adolescence are shown in Table 2. When we catcuhated the risk associated with exposure to smoke from family members on an individual hasis ('mother, fathcr, siblings, and others). there was a slight increase in risk for all exposurcx, although the incrcascs in risk were not statistically significant! How- ever. when we calculated risk according to smoke-years of exposure, which re- flects total exposure to smoke from all household members, a significantlyclu- vated risk of 2.4 (954' Cl = 1.1-5:4)was observed for womcn exposed 22 years or more. Table 2 also shows the effect of en- vironmental tobacco smoke exposure during adulthood on lung cancer risk. Women who lived for 40 or more years of their adult lives with husbands and other individual's who smoked were found to have an elevated risk of 2.4 (95% Cl = 1.1-5-3). If we considered only smoke exposure from husbands for 40 or more smoke-years (data not shown), the risk estimate decreased slightly to 2.2 (95% Cl = 1.0-4.9). In terms of total lifetime smoke-years of exposure (Table 2), no significant ex- cess risks were observed for women re- porting fewer than 40 lifetime smoke- years, but women reporting 40 or more years of exposure experienced an ele- vatedlung cancer risk of 2.3(95°k Cl = 1.1-4.6). We also examined the relationship between the lung cancer risk associated with environmental tobacco smoke ex- posure and lung cancer cell type. Since 61.4% of the lung cancers in the study were adcnocarcinomas, all lung cancer histologies other than adenocarcinoma were combined in one group for anat- ysis. Risk estimates for smoke-years of exposure were calculated'separatcly for the two groups. and the results are shown in Table 2: For women with adenocarcinoma, the risks wcre slightly elevated for all cate- gories of smoke exposure, but the re- sults did not achieve statistical significance. Women with non-adeno- carcinoma lung cancers, on the other hand, showed significantly elevated risks when their exposure to enviiron- mental tobacco smoke was of long dura- tion. The OR indicated! a threefold increased risk of lung cancer for women who reported 22 or more smoke-years of exposure from parents, siblings, and others during childhood and adoles- cence (OR = 3.4; 95% CI = 1.1-10.6). Similarly, women with 40 or more years of adulthood exposure to smoke from husband's and other holuehold members experienced a significant elevation in risk (OR = 3.3; 95% CI = 1.1-9.8). When total lifetime exposure to enr vironmental smoke was considered, the OR was 3.3 (95% Cl = 1.2-8.9) for the highest exposure level. For women with non-adcnocarcinoma: lung canccrs. thcrc was a statistically significant trend of increasing risk associated with increas- ing smoke-years of exposure for each type of exposure (childhood, adulthood, and lifetime). Since surrogate respondents were rc- yuired for about two thirds of the catic patient intcrvii:ws, we investigated whether the source of the case patient interview (sclf-respondant versus surro- gate respondent) affected the risk csli- mates. Surrogate respondents for case patients were divided into two groups. "husbands" and "other sUrrflL`atcS.- thc tatter group consisting primarily of sons andidaughtcrs. The results of this analysis are shown in Table 3. Because the number of respondents in some cate- gories was very small, analysis of risk associated with exposure to smoke from individual' household members is limited to fathers and husbands. In com- parison with the risk estimates for women whose interviews were eomi- pleted by themselves or by their hus- bands, the risk estimates for women in the "other surrogate" respondent cate- gory were considerably lower. This re- sult was true both for risk associated Tablt 2_ Effect of environmental lubacctrsmokc on lung canccr~ risk of nonsmoking women, according to mmor cell lypc All lung canccr,. Adanocarcinarrna% All other eell rcpr. keppwre himury OR' 95gf CI P for trend OR' 9S1/1 Cl P for trend OR, 95% Cl P for trend E:ry.wrc: yu~/n,r A{olhcr 1.6 41;64.3 1.6 0.5.5.4 1.7 U:}.K.2 Father 1.2 11:6.2:3 1.1 0.5-14 1.4 U:5-3.7 SnflinE% and orhen 1.7 II.K=3.9 1.3 0.6-2.7 1.4 U.6-3.7 wusbsnds 1.6 I1:K-.111 1.3 0.6-2.7 2.2 0.K-5.K Smukc-vcarz Chitdli.Kdladolccccncc (furrnn and siblings) <IK .6 1.7-3.n .8 :7-4.4 .3 .4-4.3 1K-'_1 t.l It 0:4-3.0 1.3 U..1-4.6 a, 2= 2:# 1.1.5.1 IIp 1.9 0:7-5.U .491 3.4 I.I-10:6 .U41 Adulrhwd (huwbands andorhrra <22 1.6 0.8-1? 1.7 0.8-1.7 1_S U:S4:2 :^_.t4 1.4 11.7-2.9 t.1 0.5-2.5 2.0 0:7-9.4 2040 2.J .1125 1'.8 0.7-4.7 .3211 3.3 1.1-9.8 lNli Allllifdime htwvhold cxTM+surr. <22 l.i ll:b.2 3 1.2 0.5-2.7 1.2 0.4-3.4 `_. i9 1.1 10-2:4 1.2 03-2.7. 1.5 U.h-4.2 s 4u 2.i I.1-416 011 1.7 0.8-39 .491: 3.3 V2-8.9 out •OR. adjuzted for age. race. and cducatian (Ff and>K gradrs) } Wot~: K4. No. 18. Scptcmbcr Ifi, 1992 REPORTS t419

T.trk 3. Effea of environmeaul lobaoo smoke on long anccr risk of nonsmoking women, aceording to source of tast: pslient interview $ottrceof eLK patient interview Sclf (n - 70) , Hu%band (n . 48) Other wrroEatc (n  92) Sct( .nd, huaband (n . I I>4) I:.xprsarc history OR' 95% Cl OR' 9t5i• CI OR' 95% Cl OR• 951.4 CI E:f*i+urc: ycann Fathrr Ilu.hanJ T.2 3.1 : n.4.19L,1 2a1 11.9-III.h i,I 11:4-tt'8' 011 11:7+1Z:71L4. 11.3-1.t t/.4-IL4. 2:7 3.1 LII-7:) 1 ?-NIl SihlinF, and olhur.. Cm.rkc-yean. (1ildtwwxU"&--,ccnn (par.ntk and~xiblinF.) < 12i .3 '.1-1h.7 2.4 1.4-12.R 11:6 - 1.2-1.4 .h '-1f1.X 11t-21 • 2.4 113-t1UhI.h n?-1(1.511!h 1l:2-2 1/ 2.1 /1_h-7,: i22 . h.S 1.7-'-5:5 1.4 W.'--4.h 1.6 11:5-t.K , 4.4 I .J-1:Z,> Y frir tra•nJ .11.1 79 Ni .117 AduIlMxKI (h'u.hand and atliar%) -» :=-j4 3.4 3.h 11.9-12:_'i.1' 1.11- I 1. 1 1.>: . n.7-1'4.8. /I:S' 11.4-Y:2 0.8 . 11:3•2:11 3.3 :.411 4.7 L'-14.1 4.2 /l.Y.-":4 1.5 4110.9 ~. 4:7~ 1.3~- 14~.7 Y fur trcnJ .112 .115 15 .IA'k( All Iifalimc h.Ru.cht+ld. .clMr.ur. . .-.2 2.11 (1.5.7.6 i.' 11.7-1 4.y. 1/.h u.}-1.h ?'.c u.tr7, 22-i4 4.4 1. i.15.h 1.11 0.2-5.6 11.6 11.2-I.4 3.1 I.1-K b 7111 4.1' 1.1-14.u TC tL7-1(+.7 1.5 (1.7-t,h 4.0 1.J -11. i !' hir Ir.ndi .111 IG .11 014: '(IK. adju~IrJ hn aE.:,racu. and .vlucaiim (sK and >>+ gradr.) 1420 with exposure Io smoke from individual household members and for risk associ- ated with household smoke-years of exposure. Given that the risk estimates for sel6 respondents and spouse respondents were so similar, we repeated the anal- ysis, excluding "other surrogate" re- spondents. As shown in Table 3, limiting the ease patients to self- r+espondents and spouse respondents re- sulted in significantly elevated risks as- socialed with exposure during childhood and adolescence to smoke from siblings and others (OR = 4.3; 95% C1 = 1.3-14.2) and with exposure during adulthood to smoke from husbands (OR = 3.1; 95°/'o Cl = 1.2-8.3). When smoke- years of exposure were considered, sta- tistically significant increases in risk were seen for the lowest and highest categories of smoke-years of exposure to smoke from parents and siblings; the risk estimate for the highest exposure dtegory was elevated to 4.4 (95% Cl = 1.4-135). When smoke-years of adult- hood exposure to smoke from husbands and'others were considered, statistically significant elevations in risk were seen for the lowest and highest exposure cat- egories and borderline significance was found for the intermediate category. For the highest exposure category, the risk estimate increased to 4.7 (95% Cl = 1.5-14.7). When lifetime household ex- posure was considered, a clear dosc- response effectwas evident: Risk esti- mates increased from 2.5 for fewer than 22 smoke-years of exposure to 4.(1 for 40 or more years of exposure. The sam- ple became small when we considered differences between lung cancer cell types for the different respondent. groups; nevertheless, higher estimates were still associated with women diag- nosed with non-adenocarcinoma I!ung cancers, consistent with the analysis from the full, data set. We found no statistically significant increase in risk associated with ex- posure to environmental tobaeco smoke a1 work or during social activities (data not shown). Discussion The results of our study indicate that the risk of lung cancer is increased among women who are themsclvcs life- fime nonsmokers but who live in house- holds with smokers. Elevatedrisks were seen most consistently when exposures to household smoke occurred during ad'ul'thood'. Women with non-adeno- carcinoma lung cancers who reported high levels of exposure to household smoke had the most pronounced eleva- tion in risk.. There was suggestive evi- dcnce that prolonged exposure to tobacco smoke during childhood and adolescence might also hc associated with an increased risk of lung cancer. An association between exposure to environmental tobacco smoke and the development of non-adenocarcinoma lung cancers has been noted in several other investigations, both in the United States andabroad. In a study that pooled data from, three case-control studies in the United States, Dalager et al. (9)'~ re- ported an almost threefold increased risk of squamous cell and small-ccll lung carcinomas among male and female lifetime nonsmokers who were married to smokers, hut they observed no increased risk of adenocarcinomati. Garfinkcl et al. (10) investigated the distribution of various lung cancer hiti- tologies among lifetime nonsmoking women in New Jersey and Ohio. They found thaC women whose hushand% smoked had a fivefold increase in risk of squamous cell lung carcinomas, hut no statistically significant increase in risk of other types of lung cancer. In a study conducted in Louisiana, Correa et al. (6) reported a doubling of lung can- eer risk among nonsmokers married lo smokers, both male and female. Although the risk estimates were not presented according to histologic classi- fication of the lung cancers, the authors Journal of the National Cancer Institutc 202351.3030 . . .S . . F-_Y. Y : -p4 -. .'.

indicated that the exclusion of adenocar- cinomas from their analysis produced a statistically significant linear trend of incrcasing risk with increasing ex- posure. In an investigation of non-adeno- carcinoma lung cancers in nonsmoking Athenian women, Trichopoulos e( al. (l1') reported a risk cstimatc of 2.4 for womcn whose husbands smoked fewer than 21 cigarettes per day and a risk c.ctimatc of,3.4 for women whose hus- bands smoked more than 20 cigarettes per day. An association between mar- riagc to a smoker and an incrcased' risk of small-ccll and syuamous cell lung carcinomas was also observed in a study of Swedish women (f2). For other lung cancer ccll types in the Swedish study, the risks were close to unity except in the case of women with high exposure lcvcls: whosrrixk was doubled. A study of lifetime nonsmoking women ini Hong Kong by Koo ct al. (1.3).howcd thal cx- posurc to environmental tobacco smoke was aswciatcd with an elevated risk of squamous eelll and large-cell car- cinomas. These results conflict with those observed in a multicentered study in the United States (4), where an clk- vated risk of lung canecr attributable to passive smoking was limited to patients with adenocarcinoma; these individuals experienced an almost 50% increase in risk. In another study conducted in Hong Kong, Lam et al. (14) reported that lifetime nonsmoking women with husbands who smoked had' an increased risk for the development of small-eell carcinomas, adenocarciinomas, and large-ccll carcinomas, but not squamous cell carcinomas. Our analysis of the effect of cigarette smoke exposure during childhood and' adolescence on lung'canccr risk re- vealed an increased risk in women with 22 or more total smoke-years of ex- posure. This excess risk for the highest exposure category was statistically sig- nificant for all lung cancers combined and for non-adenocarcinoma lung can- eers. A positive association between lung cancer risk and' high levels of en- vironmcntal smoke exposure at a young age has also been reported by Jancrich el al. (5); and an increased overall can- cer risk among individuals exposed to cigarette smoke in childhood has been noted' by Sandler et al. (7). In studies control subjects may have been incor- conducted by Correa et a1. (6) and Wu rectly classified as nonsmokers. et al: (8), the increased lung cancer risk In any analysis that distinguishes be- resulting from childhood exposures was tween lung cancer cell types, the ac- found to be associated specifically with curacy of the histological classification maternal smoking, whereas in our study must be considered. In the present we observed less variation in risk ac- study, which included' many hospitals cording to which family, member, throughout central Florida, an indepcnd- smokcd. In the interpretation of the results of this study, one area of particular con- cern is the impact of respondent type on the risk estimates. Alt'hough the risk estimates based on responses by the case patient and her husband are reason- ably simifar, given the samplc size, the risk estimates based on "other surrcr gatc" respondents (primarily sons and daughtcrs) are conxidcrably lower. One possible explanation is that the children of the case patients selectively under- estimated the exposures that their mothers received, particularly from their faihcrx' cigarette smoke. Convcrscly, the case patients and their husbands may have overestimated the exposure. Few studies have presented risk esti!- matcs according to respondent type. Janerich et all. (5) observed that inter- views with surrogate respondents pro- duced lower passive smoking risk estimates than direct interviews. Data on variations within the surrogate group were not available. in another study on passive smoking, Garfinkel et al. (10) reported that the highest lung cancer risk estimates were based on responses by the children of the study subjects, a finding opposite to that reported here. Another issue that must be considered in the evaluation of our findings is the possibility of misclassification of smokers as nonsmokers. Although every, attempt was made to vcrify that the case patients were truly lifetime nonsmokers, only, oral reports of smoking status were available. Fontharrt et al. (4) ascertained' urinary cotinine levels to confirm self- reports of smoking status. The results of this validation procedure suggested that, depending on the urinary eotinine level selected to indicate active smoking',, either O.K°fd'or 2.4% of case patients and 21% or 3.9% of population control sub- jects could have been misclassified as lifetime nonsmokers. If these same perr ccntages are applied to our, study, then two to five case patients and six to 12 cnt pathological, review was not poxsi- ble. However, the distr'ibution of lung cancer cell types im our study was simi- I'ar, to that desctihcd' by Fontham et al. (4), who did conduct an independent pathological review. In the latter study; the pathological review resulted in. rc- claxsificatiom of approximately 25% of syuamous, cell carcinomas and 39'/r of largc-ccil carcinomas as adcnocar- cinomas. Of tumors originally clastiificd as adcnocarcinomas, 97% were cor- rcctly classified according to the inde- pendent rcviewerti. By extrapolation. any misclassificalions in our own xtudy would tend to reduce the differences in risk estimates between various lung can- cer cell types, since the excess risks were associated primarily with' non- adcnocarcinomas. In conclusion; the results described' here suggest that long-term exposure to environmental tobacco smoke increases the risk of lung cancer in women who are nonsmokers. Risks appeared most elevated for non-adenocarcinoma lung cancers. High levels of exposure during youth and adulthood may each play a role in increasing lung cancer risk. References (/) S1111tiF.ONI GENBMAL., DFrAnTMPNT 0I . WA1.T11, EIHIt'ATH/A, ANI/ WFa.1ARF: The Health Con.cqucnccs of Involuntan- Smok- ing: A Report nf the Surgeun General. DHEW Publ No. (CDC)g7.N=9N! Wa.h- ington, DC: US Govt Print Off! 191;6 (2). BOAItU ONENVIRnNMFNTAL STIIU/F.\. AN/) ToxICOLOGY, COMMtTTE.E ON PA9sIV1. _ _ SM/twlNt, NATnoNAL RESEARC-H COru:CIL NATIONAL ACAAEMY nrSCIEN(1<s:. En% 'lr[1n- O mcntat Tobacco Smoke: Mcasuring, Ex- posurc and Aslaessing Hcalth Effcct. Wa.hington, DC:: Natl! Acad Prc-. 1HKh N CJ NATmNAI! UN>-TllurF roR OI c-tT nlln.:a ~ SAfFTiY ANnHF:ALTN, D}rAnl111iNT rrl ' ` . ingtun, DC: US Guvt Print OH: Iv91 (J), t-/1NT1/AME-r, COKMfAP. WI'-WIL1.1 \\I\./{.. eTAU Lung cancer in~nonsmokingwumcn: HVA1.TH ANIf. HIIMAN. SEKVI('EV NIOSI rt CurrenlInlethgence Bulletintl: En.irun• ~ mcntat Tobacco Smoke in the Wixkplacc. O Lung Cancer and Othcr Hralth Effect.. r~ (NIOSH)YI-IOR Wa',h W- DHHS Publ No 1"i Vul. S-t. No. 18, Scptembcr 16. 1992 2 REPORTS 1421

/ A nrallieenter ase-eontrol study. Cancer F.pidemiol„ Biomarkcrr & Prcv 1:35-43„ 1991 (5) JArvEat,r DT, Tranwr%on WD, VArrFU LR, eT AL-: Lun= anacr, .nd'cxpowrc to tobacco smokc in the houschold. NI Engl 1 Mcd 32.11:632-6.16., I9M (b) CrntKr:A P. Pnxut LW. Fr+rrnIAa+ E, r:T At.: Passive smoking and lung cancer. Lanccl 2:595-597, I9r;1 (7) SANUIX.a DP.. w.u.cti,.c: AJ. Evra,nm . RH. Cumubtivc c(fccrs of tifclimc pasivc %muk- iaE olr cancer ri..k. tancel 1:3I2-315. 1985 tN) Wu AH, l4r.wlNatrwn+ BE, Pn:r MC, r:T, At :. Srmrking and othcnrisk fucturs for lung can• ccr in wrurkn.'1NC1 74:747-75I. 11085 (V) DArwuraM NA. Pn'ra.r- I.W. M...rrn TJ, tt wu Thc rclati..n rd pa.nivc smoking lh lung ancar: Cancer ite. 4h:4tults-4rt 11. IYttH (/0/ GwaiNr:al I_ AnrwNAr-u 0. 10 Mnn-n'r l.: In- vrduntary smukinE and'lung cancer: A cax- cuntrul audy. JNCI 7t:4n5-46Y; lYKS (11) Tarrirnnrnrrrs I). KArAnutrn L. Srnwr.r+ (- r. Ar : Lung cancer anJ La..ivc vmukinE. Im J Canccr 27;1-4. IYKI ('J.) Ph:rsnAr;Ar: (i, Itntqr4a' Z• S%'4.N59EIN C: Pa..ivc r:rm+tinF and lung cancer in SwcJi.h wrwn.n. Am J Et+iilcrniol 125:17-24. 1Y7;7 (t.fl K2wr LC. I(u JH. S.w D. rrr ..I: Maa.ura- mcnl, of pas.\IVC smoking and eslimatc, uf lung cancer risk amung mrn..mukinE Chi~ ecac tcmale.. IntJ Cancer 39:1h2-t6V, 19>;7. (14) LAM TH. Ktn+r. IT, Wor:a CM. tr AI.: Smoking. pa.civc smoking and hianlogical: types in lung cancer in Hong Kong Chinc.c wonren. Br J Cancer S6'.677-G78; 1987 Second Cancers in Patients With Chronic Lymphocytic Leukemia Lois B. Travis, * Rochelle E. Curtis, Benjamin F. Hankey, Joseph F. Fraumeni, Jr. Background: Reports to date have provided widely divergent estimates of the risk of second' malignant neo- plasms in patients with chronic lym- phoeytic leukemia (CLL), ranging from cancer deficits to excesses of twofold lo threefold. Purpose: Our purpose was to estimate the risk of second primary cancers folllowing CLL, utilizing population-based tu- mor registries, and to determine whether site-specific excesses might be associated with type of ini~tial treatment for CLL. Methods: We analyzed data for 9456 patients diagnosed with CLL as a first pri- mary cancer between 1973 and 1988, who were reported to one of nine tu- mor registries participating in the National Cancer Institute's Sur- veillance, Epidemiology, and End Results (SEER) Program and who survived 2 or more months. SEER files were searched for invasive pri- mary malignancies that developed at least 2 months after the initial CLL diagnosis. Results: Compared with the general popubatiion, CLL patients demonstrated a signiificantly in- ereased risk or developing all second eancers (840 observed; observed-to- expected ratio /O/EJ', = 1.28; 95% con- fidence interval (Cll = 1.19-1 -17). Sig- nificant excesses were noted for cancers oi the lung (O/E = 1.901, hrain (O/E = 1.98), and eYe (intraocu- lar melanoma) (O/E = 3.97)' as well as malignant melanoma (O/E = 2.79) and Hodgkin's disease (O/E = 7.69). Cancer risk, which did not vary ac- cording to initial treatment category, was also constant across all time iin- tervats after CLL diagnosis. Conclu- sion: CLL patients are at a signifi- cantly increased risk of developing a second malignant neoplasm. The pat- tern of cancer excesses suggests a susceptibility state permitting the de- velopment of selected second malig- nancies in patients with CLL, perhaps hecause of shared etiologic factors, immunologic impairment, and/or quent malignancies among cancer sur- vivors (10). It is important to clarify the risk of second cancers in CLL patients because of the potential impact on pa- tient managemenl; follow-up: and sur- vival. However, various reportti Io date have provided'divcrgcntestimatct of the oceurrcncc of second maliunaneieti in CLL patients, ranging from canccri dcfi- eits to excesses of twofold to threefold (11-20). To furthcr explore and quantify the risk of second cancr:rs among a large number of C'LIL patients in the general pnpuialion and lo examine asstt- cialion. of risk with initial Iherapy. we cunduclcd a survey of mure than 4(IQIII sucli suhjccth reported Itt the Natiunal Cancer Ilnstitutr'. Surveillance. Epi- duminlugy: and End Rc.ult.(S€ER) Program' from 197_3thrriugh IyKI+. Since CLL patients are frcyuuntlj• trtalyd only with alkyl:Itine .ILCnts wilh- uul, the confounding cttcct. (tf radiu- Ihcrapy: this group of palicnts, provides a special opportunity to study the late scyuclac of these drugs. Patients and Methods We analyzed all patients diagnosed with CLL as a first primary cancer be- tween 1973 and 1988 who were reported to one of nine population-based cancer registries of the SEER Program and suo- vived 2 or more months. Such registries include those in the metropolitan areas of other influences. AI'though our results do not suggest a strong treatment effect, more detailed studies of second tumors in CLL are needed to investi= gate the role of radiation therapy and chemotherapy. (J Natll Cancer Inst 84:1422-1427,19921 Paticnts with chronic liymphocytic Icukcmia (CLL) exhibit a varicty of im- munologic perturbations (1-4) that may incrcase their risk for second malisnanl' ncoplasms. The occurrence of familial CLL may also suggest, for some sub- jects, genetic determinants (5-7). such, as those that underlie other sets of mul- tiple primary cancers (8,9): Moreover. radiotherapy and chemotherapeulic agents may also contribute to subse- Received January 10. 1442; rc. i.rd May 14. 14F)_: ar:cc(wcd June 4. IV92- tl.. US. Travi., R. E. Curtis. J. F. Fraum.•ni. Jr.. (EPiJ:mi+dngy and Iiiu.latisric. Pn+gram: Di%i- sinn nf ('anccr Efio+lnyy): H. F. Ilankcs (('ancor tirati+lic, Iltranch, SurvsillLncc Prc+pram. Dn i.iun of ('anccr Prcvcntiun andiContnA). National ('an, cir In.titutc, NkahrsJa. Md. Wc arc indebted lu thc Stalc Ilcalth RcEmn ++f Ii+wa (Ms. Kathleen McKcen and Dr. Charlc. F. Lynch) for data rctricval-to Dn. John Buic.. N.il ('af..ra.u, and Martha 'Lincr f+x criti.al rc% i.•.v of the manu.cript, and In M:•. SandA • (~ititprr.mith and, M.. Shirley Carv+n, (br v.crn:+riel %upp.+rt. 'Cr.rrrvlwnrdi-nrr ur: Inis R: l ra% i.-, M I).. En. cYuliic Plaza Ntuthi Sic. 4418. Narional In.titWr, nf l l.alth. tktheula, MII 2In{Y*_. ' k:J; nulc: Thc program is a..t ++( gr++graph- ii:alle rkrrncd, pe+(wlation.baserl ecntralUum,m,rrgi- slrie.t in the United Stalcs. o)+erntcd h. local mmpruriiorganinrrir+ns under cnnvan toth,•Nr- tiunaC Cancer Institute (NCI) Each regi.arv an- nuallrsubmit.* its casos.rn.thc NC'Lr•n a a+mpulrr talx. These computer tapc.,arcthen edited bv thc. N('1 and madc availlablc fnranalvsis. ~ Journal of thc National Canirr Ih,titulc i 1422