Tobacco Documents Online

Breast cancer and N-acetyltransferase 2 gene with currentt smoking (OR 1.00, 95% CI 092-1.09), but significant. yet very small, effects from former smoking (OR 1.10, 95% CI 1.01-1.19) (Baron et al., 1996). Again, there were no associations with smok- ing duration or intensity. An extensive review of the literature on smoking and breast cancer showed overall inconsistency, with reports of both adverse and protective effects (Palmer & Rosenberg, 1993). It is conceivable that smoking has a small protective effect against breast cancer secondary to an anti- oestrogenic action (Baron, 1984),-but this could be counterbalanced by procarcinogenic effects of tobac- co smoke constituents (Baron et al., 1990). The result may be a tow carcinogenic threshold for tobacco smoke and breast cancer that plateaus at low levels of smoking and passive exposure. The possibility of an anti-oestrogenic influence has been countered by evidence of little effect of smoking on plasma levels of oestrogen (Thomas et al., 1993), except during preg- nancy (Bernstein et a1., 1989). However, other data suggest that smoking notably induces 2-hydroxy- lation of oestradiol (Michnovicz et al., 1986, 1988). likely decreasing the bioavailability of oestrogen to breast tissue. In addition, appetite suppression by smoking could influence oestrogen metabolism through a decrease in adipose tissue mass (Hershcopf & Bradlow, 1987). Smoking has also been associated with earlier age at menopause (Hiatt & Fireman. 1986), which could alone lower risk. We tested this in the postmenopausal group and found, on average, current smokers had an earlier age at menopause than never-smokers (45 versus 48, P< 0.1). Another case-control study found positive associa- tions with breast cancer risk for passive smoking but not ever-active smoking (Smith et al., 1994). Two other recent studies that used referent groups with no active or passive smoke exposure reported signifi- cant positive associations between breast cancer risk and passive exposure (OR of around 2-3) but, surprisingly, similar ORs (around 2-4) for current and former active smoking (Morabia et al., 1996: Lash & Aschengrau, 1999). Lash & Aschengrau (1999) found former smokers and subjects exposed to passive smoke before the age of 12 years had the highest risk (significant ORs between 4 and 8). Given these reports, it is conceivable that there is a non- linear exposure-response relationship with a low threshold. This could explain why epidemiological studies, including the present, have generally found little to no evidence of an association between breast cancer and smoking. Compared with common case-coritrol designs, the present case-control design has the potential to reduce participation. recall and interviewer biases because eligible patients. participants and inter- 467 viewers, respectively, are unaware of case-control status. Controls undergo breast cancer detection similar to cases. and are selected under similar conditions. Interviewing patients prior to diagnosis minimizes or eliminates both the physiological effects of treatment and any influences of health-related information on the perception of lifestyle behaviours such as smoking. Also. the present design shares advantages of incident over prevalent case-control studies in that all diagnosed cases are invited to participate, regardless of duration of disease or treat- ment success, and the recall of events and habits prior to diagnosis is enhanced. Most of the patients in this study were recruited from breast centres serving largely white and well- educated women. Therefore, the present findings may not be externally valid for other populations of poorer women or women in other racial or ethnic groups. Also, the control group may not be representative of women who do not have breast cancer because women who are biopsied may be over concentrated with the 'worried well', i.e. women who get screened more than other women. To test this possibility. we examined case versus control responses to questions on the frequency of past mammographic screening in women under 51 years and over 50 years old. In younger women, there was a small difference in the proportion of subjects who were screened at least every 2 years (cases 72%, controls 78%) or every year (cases 44%, controls 53%). There was a some- what greater disparity in older women who were screened at least every 2 years (cases 70%. controls 82%) or every year (cases 60%. controls 72%). These modest differences may be a reflection of the worried- well phenomenon that could bias risk factor associa- tions. However, the expectation for smoking effects is that worried-well controls would be less likely to smoke than cases, but in fact, ORs for smoking variables were nearly all -_ 1.0. It is possible that NAT2 is a risk factor for benign breast disease in the same manner that it is a risk factor for breast cancer. As with over-matched cases and controls, this could have biased findings towards the null hypothesis. The only study to compare acetylator phenotype between benign breast disease and healthy controls found nonsignificant differences for cystic disease with epithelial hyperplasia versus normal controls (43.8% versus 55.2% slow acetyla- tors. respectively) and for 'cystic disease' alone (42.6 versus 55?% slow acetylators. respectively) but no differences for 'fibroadenoma' alone (Philip et al., 1987). In the present study. approximately half of controls were slow acetylators, as expected in Cauca- sian populations (Lin et al.. 1993). It is also possible that smoking is a positive risk factor for both benign

Breast cancer and N-acetyltransferase 2 gene frequency was 64%. and one of two African Amer- icans was a slow acetylator. Regression models were retested with white subjects alone and the results were not altered. Therefore, the models presented include all subjects. There were no significant effects of NAT2 genotype on breast cancer risk using all controls or restricting the analysis to lower or higher risk benign breast disease controls (Table 2). There was also no evi- dence of a difference in the distribution of NAT2 by cancer histopathology (data not shown). The propor- tion of slow NAT2 by histology was 61% for 72 invasive ductal carcinomas, 64% for 11 invasive lobular carcinomas, and 63% for CIS. In age-adjusted logistic regression models, there was no significant risk of invasive cancers from slow versus fast NAT2 (OR 1.25, 95% CI 0.73-2.13). Table 3 shows multivariate-adjusted effects of the smoking variables. The majority of cases were never- smokers (59%), with approximately half them having low passive exposure as well. There was no signifi- cant increase in risk of breast cancer among active former or current smokers as compared with the, referent group with no active or passive exposure. Also, there was no significant risk associated with passive smoking status among never-smokers. There was no association of breast cancer with either dura- tion of smoking or number of cigarettes smoked per day. Most odds ratios are _ 1. The relationship of cancer risk to active smoking was re-examined for 465 invasive cancer cases alone, but the results were unchanged (not shown). There was no multiplicative interaction using pro- duct terms for any combination of NAT2 genotype and active or passive smoking (P>0.3). All the above modeLs were retested separately in pre- and postmenopausal women. There was no association of breast cancer with NAT2 in either pre- or postmeno- pausal women (OR for slow NAT2 1.15, 95% CI. 0.49-2.77: OR 1.29, 95% Cl. 0.74-2.27, respec- tively). In both pre- and postmenopausal groups, breast cancer risk was not associated with active smoking. Parameter estimates were negative or close to 0. Although not statistically significant, the mag- nitude of estimates for passive smoke exposure in never-smokers differed between the two groups: pre- menopausal OR 2.69, 95% CI. 0.91-8.00; postmeno- pausal OR 1.01 95% CI. 0.45-2.27. However- this was limited by a small sample size of only 21 cancer cases among premenopausal never-smokers. Multi- plicative interaction models in pre- and postmeno- pausal women gave no evidence that passive or active smoking effects differed by NAT2 genotype. Discussion We found no independent associations between NAT2 and breast cancer risk. There were no signifi- cant associations with active or passive smoking, and ORs were generally < 1.0. Among never-smokers, Table 2. N-acetyltransferase 2 (NAT2) genetic polymorphisms and risk of breast cancer Cases no. (%) All controls° no_ (%) OR (95% CI)^ Low-risk controls` no. (%) OR (95% CI) High-risk controls" no. (46) OR (95% CI) Toml no. (%) Fast acetylators 43 (38) 120 (43) 1.00 (referent) 48 (45) 1.00 (referent) 61 (41) 1.00 (referent) 163 (42) wT/WT 3 7 23 8 13 30 wT/T 41c i90 26 56 25 24 82 WTC A 10 38 14 22 48 WT/GBS7A 0 3 1 2 3 Slow acetylators ) 3 70 (62) 158 (57) 3.25 (0-78-2.00) 59 (55) 135 (0.76-2.41) 87 (59) 1_12 (0.66-7.89) 228 (58) T 4 C/T391c 18 43 19 22 61 T391C/C'9UA 38 88 31 5] 126 T391C/G857A 3 5 1 2 8 G590A/G590A 9 18 7 10 27 G590A/G857 A 1 2 0 2 3 G857 A/G857 A 1 2 1 0 3 WT. wild-type allele lacking the slow-acetylator mutations T34iC. G590A. G6i A. 'Includes 23 women with insufficient tissue from benign breast disease biopsy to,classify the degree of cellular proliferation For low and high risk classification. bFrom unconditional logistic regression models adjusted for age and menopausal status. °Normal breast or benign breast disease histopathologies with nonproliferative changes. dBenign breast disease histopathologies of hyperplasia with no atypia. atypical hyperplasia or complex fibroadenomas (sclerosing adenosis, intraductal papilloma). OR, odds ratio: CI, confidence interval.

Phnrmacogertetics 2000, ]0:467-469 Breast cancer, passive and active cigarette smoking and N- acetyltransferase 2 genotype Ralph J. Delfinoa, Cynthia Smitha, John G. Westh, Henry J. Lin`, Ldavard 1Vhited. Shu-Yuan Liaoa•e, Jason S.Y. Gim°, lloang L. f44a°, John Butlerr and Hoda Anton-Culvera "Epidentmtogy Division. Department of Medtcine. University of Califorrtfa. Irvine. California. "Ereact Cnre Center o(Orangc. Orange. California, `Division of Medical Genetics. Harbor-University of Catifornia. Los Angeles. Medical Center. Torrance. CaLfoialia dSaddteback Breast Center. Laguna Hilts. California. `Departnlen[ of Pathology. College of Aledlciree. Unlversity of California. irvine. California and rDepartment of Surgery. College of Medicine. University of Ca7ifornia. Irrine. California. USA Received 15 September 1999; accepted 27 November 1999 The relationship of breast cancer to cigarette smoking is inconsistent in the literature, possibly due in part to heterogeneity in carcinogen metabolism. N-acetyltransferase 2 (NAT2) enzyme activity is believed to play a role in the activation of tobacco smoke carcinogens. We examined the effect of NAT2 genetic polymorphisms on risk of breast cancer from active and passive smoking. Women were recruited from those who had suspicious breast masses detected clinically and/or mammographically- Questionnaire data were collected prior to biopsy diagnosis to blind subjects and interviewers. Histopathology showed I13 cases with mammary carcinoma (30 carcinoma in situ) and 278 controls with benign breast disease. NAT2 genotype was determined using allele-specific polymerase chain reaction amplification to detect slow acetylator mutations. Effects of passive and active tobacco smoke and of NAT2 genotype on breast cancer risk were exanrined with logistic regression controlling for known risk factors. Models first included all controls, and subsequently 107 with no or low risk (normal breast or no hyperplasia), and finally 148 with high risk (hyperplasia, atypical hyperplasia, complex fibroadenomas). Referents had no active or passive smoke exposure. We found no association between breast cancer risk and NAT2, smoking status (never, former, current), smoking duration, or cigarettes per day. There were no effects of passive exposure among never-smokers. Models were unchanged across control groups. There were no statistical interactions between tobacco smoke exposure and NAT2. The results were similar when restricting the analysis to invasive cancers. These findings do not support the hypothesis that NAT2 is a risk factor for breast cancer or that it alters susceptibility to tobacco smoke. Pharmacogenerca to461-1e9 rr; 2000 uppinwn Williams & wilktns Keywords: breast cancer, benign breast disease, smoking, NAT2 gene Introduction Aetiological research on lifestyle factors has the potential to identify feasible approaches to preventive intervention of breast cancer. Cigarette smoking may be one such factor. However, epidemiological re- ;earch has been inconsistent (Palmer & Rosenberg. 1993). One group of promutagens and procarcino- gens in tobacco smoke are aromatic amines, includ- ing 4-aminobiphenyl. (3-naphthylam)ne. and 2- `orrespondence to Ralph I. De@ino. Epidemiology Divisinn. Department ol Medkine. Universipol California. Inine. CA 92697-7550. USA TeL -1 949 824 7401. (ar. +1 949 824 4773: e-mail. rdelfino~duci.edu Original article amino-3-methylimidazo[4.5-Qquinoline (IQ) (Yama- shita et al.. 1986; Talaska et al.. 1991)_ Adducts of IQ and 4-aminobiphenyl with DNA occur in cultured human mammary epithelial cells (Pfau et a1.. 1992; Swaminathan et aL. 1994), indicating that metabolic activation of aromatic arnines may occur in breast tissue. Polycyclic aromaticc hydrocarbons such as benzo(a)pyrene are other smoking-related carcino- gens that form DNA adducts in human breast tissue (Li et al., 1996). One possible difference in suscept- ibility to tobacco carcinogens is in the capacity of metabolic enzymes to activate or deactivate carcino- gens in their ability to bind to DNA- U960-314X t 200U[.ippincon WiltiamsS Wilkins

Breast cancer and N-acetyltransferase 2 gene carcinoma in situ (CIS). Because CIS has a known malignant potential for developing invasive cancer and is treated clinically as such. subjects with CIS were used as cases in the analysis. Pathological . classifications of benign. breast disease were made according to the criteria of Page &.Rogers (1992).. When sufficient parenchyma adjacent to lesions were available, classifications included nonproliferative dis- ease, proliferative disease withoutatypia, and atypi- cal hyperplasia (Dupont et al., 1993). These histological features were combined with fibroadeno- ma presence. which was subdivided based upon find- ing either noncomplex or complex lesions (sclerosing adenosis or intraductal papilloma). These classifica- tions allowed. us to stratify the control groups on low versus high.risk of breast cancer. In several cohort studies, elevated relative risks for subsequent develop- ment of breast cancer have been consistently identi- fied for women with atypical hyperplasia [relative risk (RR) 2.5-7.3] and women with proliferative disease without atypia (RR 1.6-3.5) (Carter et al., 1988; London et al.. 1992; Bodian et af.. 1993; Dupont et al.. 1993. 1994). Low or no risks have been associated in these studies with women without proliferative disease (RR 1.0-1.6). The presence of sclerosing adenosis or intraductal papilloma may confer additional risks (RR 3.10-23.6) (Kriegef & Hiatt, 1992; Dupont et al.. 1994). It is possible that the occurrence of some benign breast disease is due in part to risk factors shared with breast cancer cases, potentially including NAT2 or smoking. There- fore, separate analyses are reported using three control groups: all controls (n = 278), low-risk con- trols (15 normal breast. 92 no hyperplasia, no atypia), and high-risk controls (57 hyperplasia but no atypia. 27 atypical hyperplasia. 39 sclerosing adenosis and 25 intraductal papilloma). For 23 controls undergoing only fine-needle aspirations or core biopsies. there was insufficient tissue surround- ing the fibroadenoma to classify the proliferative state of the lesion. Therefore. they were used only in the analyses including all controls. NAT2 genotyping The NAT2 genotype was determined using allele- specific polymerase chain reaction (PCR) amplifica- tion (Lin et al., 1993, 1994) on DNA extracted from frozen buffy coat samples (Bell et a1., 1981: Geever et al.. 1981). For the present analysis, individuals who were either homozygous or heterozygous for wild- type NAT2 were classified as fast acetylators, whereas those carrying two slow-acetylator inutations were classified as slow acetylators. Three NAT2 mutations were assaved for each subject (T391C. NAT2*5: G'yoA. NAT2*6: and 463 Gss7A. NAT2*7) (Vatsis et a1.. 1995). There were only two black subjects in the study. One was classified as a slow acetylator by the above muta- tions. The other was tested for a fourth slow-acet- ylator mutation found virtually only among blacks (G791A, NAT2*14) (Bell et al.. 1993), and she was found to have only the wild-type nucleotide. Negative controls were included in every PCR run. A 100% duplicate sample rate was used. The laboratory team was blinded to case-control and exposure status. Statistical analysis Smoking variables included active or passive smoking status (current or former smokers, and never-smokers with high or low passive exposure). smoking dura- tion, and average cigarettes per day. Risk from active smoking was analysed as a categorical variable based upon the quartile distribution of smoking in all controls. Never-smokers with low passive exposure were the referent category. Effects of passive smoke exposure among never-smokers were analysed as a binary variable. Passive exposure was considered high if subjects reported that as an adult the people with whom they had lived smoked in their home either usually or some of the time. Passive exposure was considered low if this rarely or never occurred, or they always lived alone. The effects of NAT2 genotype and passive or active smoking on the risk of breast cancer were examined in unconditional logistic regression models. P< 0.05 was considered statistically significant for two-sided t- statistics. Multivariate analyses were used to estimate odds ratios (OR) and 95% confidence intervals (CI), and included an examination of known or suspected risk factors that may confound and/or modify the effects of interest. These included age at diagnosis, age at menarche, menopausal status, age at first full- term pregnancy, parity, total months of pregnancy, lactation history (months breast feeding), education level, race/ethnicity. family history of breast cancer in first- and second-degree relatives, and body mass index (kg weight/m2 height). Subjects were consid- ered postmenopausal if they reported cessation of menstruation > 6 months ago and were > 49 years old. Subjects less than 50 years old were considered postmenopausal if they reported natural menopause or bilateral oophorectomy. The effects of tobacco smoke exposure and NAT2 genotype on the risk of breast cancer were first assessed separately in bivariate models and then after controlling for confounding variables. Confounding was assumed if the parameter estimate for NAT2 or smoking variables changed by at least 10%. The regression analysis first involved the entire control population and. subsequently, the low-risk and high-

462 An irnportant component in !he metabolic path- way of aromatic amine carcinoe,ens is ,N'-acetpltrans- ferase (NAT) enzyme activity. NAT transfers acetyl- coenzyme A to the amino (or h,; droxyl) side chain of arylamines, converting them to unstable electro- philes. Aromatic amines such es 4-aminobiphenyl are detoxified by NAT enzymes as evidenced by higher levels of haemoglohin adducts among smokers who are slow N-acetylators as <:o:npared with fast N- acetylators (Vineis et a1., 199? i. Hererncyclic aro- matic amines, on the other hand, require enzymatic activation by NAT enzymes to electrophiles in order to bind to DNA and thus iniiiate carcinogenesis (Hein, 1988). NAT activity ha-- been linked to two genes, referred to as NAT1 and NAT2 (Blum et al., 1990; Ebisawa et al., 1991; Grant et a1., 1991: Vatsis et aI., 1991). The NAT2 gene is polymorphic. and individuals who carry two allelic mutations have a slow-acetylator phenotype, whereas heterozygous wild-type genotypes have an intermediate acetylator phenotype and homozygous wild-type genotypes have a rapid-acetylator phenotvpe (Vatsis et al.. 1991). Epidemiological research on the relationship of breast cancer risk to interactions between NAT2 genetic polymorphisms and tobacco smoking is in- consistent. One study showed an increased risk of breast cancer from smoking in slow acetylators (Ambrosone et at., 1996), while two others (Hunter et a(., 1997: Millikan et al., 1998) .showed some limited smoking effects among fast acetylators but no dose-response relationships_ There is no research in the breast cancer literature examining passive expo- sure to tobacco smoke and NAT2 polymorphisms. The objective of the present investigation was to examine the effect of polymorphisms of the NAT2 gene on the risk of breast cancer from both active and passive smoking in 113 cases with mammary carcinoma compared with 278 controls with benign breast disease. Materials and methods Design and population A case-control study was conducted on women over 39 years of age with a suspicious breast mass detected clinically and/or by diagnostic mammog- raphy, who were scheduled for an open. core or fine- needle breast biopsy to rule out mammary carcino- ma. Other eligibility criteria included no previous history of cancer, no other severe debilitating medical illnesses. and fluent spoken English. Limiting the age of eligibility to 40 years and above was intended to achieve a cancer- to benign-brcast-disease ratio of llelfino et al. = 25" (13al.cr et ril.. 199 5). Recruited patients either had a moderate ro high clinical suspicion of a breast carcinoma frorn mammography or there existed sulTicient clinical suspicion to warrant a diagnostic biopsy. Eligible women were quickly identified, re- cruited and interviewed by a trained interviewer who was present ai one of three participating breast centres in Orange County: (1) tlie Breast Care Center of Orange; (2) the Saddleback Breast Center; and (3) UCI Clinical Canc-er Center. Department of Surgerv . Prior to the biops~date, self-administered risk factor questionnaires were distributed to 374 out of 391 subjects in the present analysis. They were instructed to return questionnaires by mail. We confirmed that questionnaires were completed before subjects re- ceived their diagnosis. Seventeen subjects completed their questionnaires in the breast center while they waited for diagnostic results. Cases with malignant tumours and controls with benign masses were subsequently identified histo- pathologically. The aim of this modified case-control design was to reduce participation, recall and inter- viewer biases, which may occur in case-control stud- ies because subjects, and often interviewers. are aware of the disease status. Future studies are planned to test these expectations. Out of 535 patients approached and eligible, 391 participated fully and are included in the preseni analysis (113 cases. 278 controls), and 86 declined participation. Fifty-five other participants were ex- cluded from analysis because they failed to complete all questionnaires prior to receiving diagnostic results ( n= 51) or refused blood draw (n = 4). Three sub- jects were excluded because of missing smoking data. Two other subjects are included but only gave smoking status. Six never-smokers gave no data on passive exposure and were excluded from passive exposure analyses. The participation rate among the 535 women was similar in patients diagnosed with (82%) or without cancer (85%). The institutional review boards of the University of California. Irvine. and the Long Beach Memorial Hospital (for the Saddleback Breast Center) approved the study proto- cols in accord with an assurance filed with anii approved by the US Department of Health and Hu- man Services. Inlbrmed written consent was obtained from all subjects. Histopathologicat criteria for case and control group Crassifications The study pathologist (S-YL) conducted a blinded revicw of the histology slides for patients with benign and malignant diag,noses_ Among the cancers, there were 72 invasive ductal careinomas. 11 invasive lobular carcinomas. 28 ductal and two with lobular 2505587218

464 risk control groups. To assess the possibility that the effect of passive or active smoking differs in fast versus slow acetylators, we assessed multiplicative interactions between NAT2 genotype and active or passive smoking in adjusted logistic regression mod- els. Sample sizes were considered too small to assess additive interaction across differcnt NAT2 and smok- ing strata. The fit of the variou, rnodels was assessed with the likelihood ratio test_ Results Subject characteristics Cases were significantly older than controls (Table 1). Postmenopausal women were more likely to be cases than premenopausal women. There were no signifi- cant case-control differences for anv of the other reproductive factors. Later age at tirst full-term preg- nancy was significant only for the age category 25- Delfno et al. 29 years using low-risk controls, but older ages and nulliparity did not increase risk. There was no case- control difference in body mass index. A majority of both cases and controls were educated beyond high school, and there were no significant differences. There was a suggestion of increased risk of breast cancer in women with a family history of breast cancer in the mother or a sister, but this was nonsignificant 1 P= 0.18). There was a significant increased risk of breast cancer with a positive familv history in a second-degree relative (grandmother or aunt). Breast cancer risk fronn NAT2 genotype and smoking The population was predominantly white non-Hispa- nic (92%). The frequency of slow-acetylator muta- tions was less in the 1 5 Asian subjects (33%) than in the 360 non-Hispanic white subjects (59%), which is consistent with population estimates of acetvlator status (Lin et a!.. 1993). Am'ong 14 Hispanics, th( Table 1. Demographic and reproductive characteristics in case and control subjects Characteristic Cases (n = 113) All controls (n = 278) P' Loiv-risk controls P High-risk controls P (n=1071 (n=148) Age (years. mean. SD) 61 (12) 54 (101 0.0001 72 (10) 0.0001 54 (10) 0.0001 Menopausal status (no.. %): Premenopausal (reQ 28 (25) 113 (41) 53 (50) 49 (33) Postmenopausal 85 (75) 165 (59) 0.004 54 (50) 0.0002 99 (67) 0.1 D Age at menarche (years, mean. SDi 13 (]..3) 13 (15) 0.65 13 (1.'.i1 0.63 13 (1.5) 0.42 Age at first full-term pregnancy (nu.. °'o) < 25 years old (ref) 45 (40) 125 (45) 46 (43j 69 (46) 25-29 years old 30 (26) 47(17) 0.10 13 (12) 0.04 29 (20) 0.23 > 29 years old 11 (10) 40 (14) 0.60 22 (21) 0.41 16(11) 0.93 Nulliparous 27 (24) 66 (24) 0.77 26 (24) 0.89 34 (23) 0.70 Parity (no. and %) Nutliparous (ref) 27 (24) 66 (24) 26 (24) 34 (23) 1-2 children 54 (48) 144 (52) 0.93 58 (54) 0.87 76 (51) 0.94 > 2 children 32 (28) 68 (24) 0.97 23 (22) 0.72 38 (26) 0.89 Months pregnancy (mean, SD) ] 9 (14) 18 (14) 0.93 16 (12) Q47 18 (15) 0.67 Months breast feeding (mean, SD) 2_9 (5.4) 3.6 (6-3) 0.24 33 (6.0; 0.75 3-8 (6-8) 0.15 Body mass index (kg/m2. mean, SD~ 251 (5_0) 24.9 U.3) 0.74 24.9 (6.21 0.92 25.0 (4-8) 0.86 Education (no. %) High school or less (ref) 35 (22) 43 (15) 16 (15) 26 (17) College or vocational 60 (53) 172 (62) 0.30 65 (611 0-42 90 (61) 0.77 Postgraduate or professional 28 (25) 63 (23) 0.71 26 (24) 0.86 32 (22) 0.34 Family history of breast cancer (no_ 'Sr,)b None (ref) 64 (57) 180 (66) 70 (671 92 (63) Mother or sister 23 (21) 47 (17) 0.18 19 (18( 0_28 27 (19) 0.39 Grandmother or aunt .. 25 (22) 46 (17) 0.02 16 (15) 0.04 26 (18) 0.11 aTwo-sided P-values for continuous variables are from Wilcoxon rank-sum tests. and for categorical variables they arc from unconditional logistic regression as compared to referent categories (rel). adjusted for age. Menopausal status is n,,: adjusted for age. eThe family history of one case and five controls is unknown. 2505587220

466 Table 3. Active and passive cigarctte smoking and risk of breast cancer Delfino ct al. .Smokinp variable Cases cmitrols° OR (95°k C1J^ Lou-risk conlrols` OR (9S%Cl) High-risk corttrots^ OR (95% C1) Smoking exposure No active/passive 33 9h 1.00 (referent) 44 ].00 (referent) 44 1.00 (referent) Passive only 31 >] " 1.32 (0.69-252) 16 1.78 (0.; 7-4-11) 30 1.03 (0.50-2.1 1; Former 40 99 0.94 (0.53-1.68) 37 0.91 (0.44-1.90) ~7 0.77 (0.40-1-47) Current 5 24 0S5 (0.18-1.67) 5 1.25 (0.27-5.82) 14 0.45 ((].14-L47) Duration smokin g Never and no ETS 33 96 1.00 (referent) 44 1.00 (re.lerent) 44 1.00 (referent) < 13 years 14 42 0.94 (0.43-2.03) 14 1.12 (0.42-3.00) 26 0.71 (0.30-1.66) 13-26 years 10 42 0.70 (.0.30-1-62) 15 0,6 3 (0.21-1.92) 22 0.59 (Q23-1.481 > 26 years 20 3S 0.74 (0.34-1.611 13 0_50 (0.16-1.52) 23 0.63 (226-1_49) Cigarettes per day None 33 96 1.00 (referent) 44 1.00 (referent) 44 1.00 (referent) < 8 per day 19 4; 1.04 (0.50-2-13) 11 1-46 (0.53-3.99) 30 0.77 (0.35-1.69) 8-25 per day 18 46 0.75 (0.35-1.58) 16 Q66 (0.24-1.83) 27 0.60 (0.26-1.39) > 25 per day 7 3 1 0.51 (0.19-1.35) 15 0-32 (0.10-1.06) 14 0.48 (O.] 6-1.45) ETS in all never-smokerse Low 33 96 1.00 (referent) 44 1.00 (referent) 44 1.00 (referent) High 31 D 1 1.50 (0.79-2.87) 16 1-86 (0.81-4.27) 30 1.20 (0.58-2.50) elncludes 23 women with insufficient tissue from benign breast disease biopsy to classify the degree of cellular proliferation for low- and high-risk classiflcation. bFrom unconditional logistic regression models adjusted for age, menopausal status. and family his;nry of breast cancer in first- and second-degree relatives- `Normal breast or benign breast disease histopathologies with nonproliferative changes. d6enign breast disease histopathologies of hyperplasia with no atypia, atypical hyperplasia or complex fibroadenomas (sclerosing adenosis, intraductal papilloma). enigh, as an adult the people with whom they had lived smoked in their home either usually or some of the time; low, as an adult tha• people with whom they had lived rarely or never smoked. or they always lived alone. OR, odds ratio; Cl. con8dence interval; ETS. environmental tobacco smoke exposure. there was a nonsignificant elTect of passive smoke exposure in premenopausal never-smokers (OR 2.69, 95% Cl, 0.91-8.00), however, the sample size was very small (21 cases). In regression models for multi- plicative interaction, there was no suggestion that NAT2 genotype modified risk from active or passive tobacco exposure. The association between arylamine exposure and bladder cancer has been well established, and risk is increased in individuals with the slowc-acetylator phenotype (Hein. 1988) and genotype (Risch et al., 1995). However, breast cancer studies have not yielded convincing evidence for Cffect modification by acetylator status. Ambrosone et a.'. (1996) conducted a case-control study and found that postmenopausal women with slow-acetylator NAT2 genotypes had an increased risk of breast cancer if they smoked (OR for > 20 cigarettes/day: 4.4, 95% CI 1.3-14.8). Neither NAT2 nor smoking was independently associated with risk, and there were no nertable associations in premenopausal women. In an analysis of the Nurses Health Study cohort, there was some suggestion of a small risk to fast acetylators who smoked prior to their first pregnancy, but no duration-response rela- tionship was observed (Hunter et al., 1997). In a case--control study by Millikan et al. (1998), there was an isolated association between breast cancer risk and being a former smoker among fast acetyla- tors who quit 1-3 years prior to interview (OR 7.4. 9 5% Ci, 1.6-32.6) and 4-9 years prior to interview (OR 4.1., 95% Cl 1_3-13.]). Similar to their results for NAT2. thev found an increased risk from NAT1 * 10 among former smokers. However, there was no effect of duration (years smoked) or intensitk of smoking (cigarettes/day) across either NAT] or NAT2 genotypes. The authors also found no multi- plicative interaction between NAT2 and smoking. None of the studies examining NAT2 genetic poly- morphisms found itt to be independently associated with breast cancer risk (Agundez et al., 1995; Ambrosone et al., 1996; Hunter et a1., 1997: Millikan et a1.. 1998). which is consistent with our overall findings. Similar to Hunter et al. (1997) and Millikan et a1. (1998), the largest case-control study to date on breast cancer and smoking also found no association 2505587222

468 breast disease and breast cancer. thus biasing find- ings towards the null hypothesis. However. in gen- eral, we found little to no difference in estimates of effect for both NAT2 and aclive smoking in the analysis involving low- versus high-risk benign breast disease controls. The literature does not sup- port an adverse effect of active-smoking on benign breast disease. Smokers have been found to be at reduced risk of benign breast disease in several stud- ies (Berkowitz et al., 1985: I'astides et al.. 1987; Wyshak et al.. 1988)_ Other studies have found no association (Nomura et a1., 1977: Mant et a1., 1986; Baron, 1987). including one case~control study restricted to cases defined as having higher-risk proliferative benign lesions (Rohan et al.. 1989). Negative bias in our evaluation of the effect of passive smoke could have occurred because we did not assess passive exposure outside of the home. Exposure misclassification is possible for women with high exposures at work or elsewhere. The present study lacked sufficient statistical power to detect a small magnitude of ~\- AT2-smoking inter- action. Evidence in this and other studies examining breast cancer. NAT2 and smoking (Ambrosone et a1., 1996; Hunter et a1., 1997; Millikan et al., 1998) suggests that if there are adverse effects of smoking, such risks are likely limited to a small group of susceptible women, thus yielding low overall risk estimates. It is likely that polymorphisms of several metabolic genes involved in carcinogen metabolism are required to increase the risk of smoking. In conclusion. the present findings do not support the hypothesis that NAT2 is a risk factor for breast cancer or that it alters susceptibility to tobacco smoke. The biological plausibility of an adverse effect of smoking on breast cancer risk remains, but the data from human populations are inconsistent. The reason may be that opposing anti-oestrogenic and carcinogenic mechanisms lead to nonlinear relation- ships that are difficult to detect. This suggests the need for expanded epidemiological studies to assess the shape of the exposure-response relationship be- tween tobacco smoke exposure and breast cancer risk. This may require the use of multiple biomarkers of susceptibility. including genotypes associated with activation and deactivation of tobacco carcinogens and with oestrogen metabolism. Acknowledgements This research was supported bv funds from the California Breast Cancer Research Program of the University of California. grant number IRB-0295. H.J.L. was supported by NIH grants 1R01CA66782 and 1R01CA73403. We wish :o thank our research Delhno et al. assistants for their diligent efforts at data collection and processing: Irene Masunaka. Randy Williams. Presley Ha- ves, and Tara McKittrick, Epidemiology Division. Department Medicine, University of Califor- nia, Irvine. We thank Dr John A. Baron. Dartmouth Medical School, for helpful comments on the manu- script. References Agundez [.A. Ladero JM. Olivera M. Abildua R. Roman JM. Benitez J. Genetic analysis of the arylamine .ti-acetyltransfer- ase polvmorphism in breast cancer patients. Oncology 199_~: 52:,'-i1. Ambrosone CB. F'reudenheim JL. Graham S. Marshall JR. Vena JE. Brasure JR, et n1. Cigarette smoking. N-acet_vltransferase 2 genetic polymorphisms. and breast cancer risk. /AM.9 1996: 276:14941501. Baker I.H. Chin DY. Wagner KV. Progress in screening for early breast cane:er. ) Surg Oncol 1995: 30:96--102. Baron JA. Smoking and estrogen-related disease. Ani 1 Epfderrvnl 19_ 84, 119:9-22- Baron IA. Smoking anc estrogen-related disease in w•omen. In: Rosenberg !vIJ. ed. Smoking and reproductfve health- Littleton. MA:PSG:1987.pp 149-166. Baron JA, La vecchla C, Levi F. The antiestrogenic effect of cigarette smoking in women. Am J Obstet Gyrtecol 1990: 162 . 502--514. Baron IA, Newcomb PA, Longnecker MP, Mittendorf R. Storer BE, Clapp RW. er a1- Cigarette smoking and breast cancer. Cancer Eptdemiol Binmark Prev 1996: 5:399-403. Bell_GI. K.aram JH. Rutter WJ. Polymorphic DNA region adjacent to the 5' end of the human insulin gene- Proc Natl Acad Sd LISA 1981: 78:57>9-5763. Bell DA. Taylor JA. Butler MA, Stephens EA. Wiest 1, Brubaker LH. et a7. Genotype/phenotype discordance [or human arvlamine N-acetyll.ransferase (NAT2) reveals a new slow- acetylator allele common in African-Americans. Carcinogen- esis 1993: 14:1689-1692. Berkowita GS. Canny PF. Vivolsi VA, Merino MJ, O'Connor TZ. Kelsey JL. Cigarette smoking and benign breast disease- 1 EpFdemial Contmtut Health 1985: 39:308-313. Bernstein L. f'ike MC. Lobo RA, Depue RH. Ross RK, Henderson BE. Cigarette smoking in pregnancy results in marked decrease in maternal hCG and oestradiol levels. Br J Obstet Gynaecol 1989: 96:92-96. Blum M. Grant DM, McBride OW, Heim M. Meyer UA. Human andamine V-acel:yltransferase genesisolations chromosomal localization and functional expression. DNA CeG Biol 1990: 9:193-203- Bodian CA. Perzin KH, Lattes R. HolTmann P. Abernathy TG. Prognostic significance of benign proliferative breast disease. Cancer 1993: 71.3896-3907. Caner C.. Corle DK, Micozzi MS. Schalzkin A, Taylor PR. A prospective study of the development of breast cancer in 16 692 women with benign breast disease. Am 1 Epidemio7 1988: 128:467-9 77. Dupont 11'D. Parl FF. I-fartmann WH. Brinton LA. Winfield AC. l1'orretl JA. et at- Breast cancer risk associated with proliferaive breast disease and atypical hyperplasia. Cancer 1993:71:1258-1265. Dupont laD. Page DL. Parl FF. Vnencak-Jones CL. Plummer WD. Rados MS. Schuyler PA. Long-term risk of breast cancer in 2505587224

Br-east cancer and N-acetyltransjerase 2 gene women with fibroadenorna. N Engl I Mcrl 1994: 331:10-1>. Ebisaaa 7'. Deguchi T. Structure and restnction fragme.nt length polymorphism of genes for human liver arylamine ,'s'- acetyltransferases Biochcm Siopkys Res Conimun 1991: 177:1252-1257. Geevcr RF, Wilson LB, 1.'allaseth FS. Milner PE. BiLtner A1, Wilson /T. Direct identification of sickle cell anemia bv blot hybridization. Proc Nat1 Arnd Sci USA 1981: 78u081- 5085. Grant DM, Blum M. Beer M. Meyer UA. Moncunorphic anei polymorphic human arylamine N-acetyltransferases: a com- parison of liver isozymes and expressed products of nvo cloned genes- Mo1 Phannacol 1991: 39:184-191. liein DW. Acetylator genntype and arylamine-induced carcino- genesis. Biochirn Biophys Acm 1988: 948r37-66. Hershcopf RJ, Bradlow HL- Obesity, diet. endogenous estrogens and the risk of hormone-sensitive cancers. Am j Clin Ntur 1987:45:283-289. Hiatt RA. Fireman BH. Smoking, menopause. and breast cancer. I,tiatl Cancer Inst 1986: 76:833-838. Hunter DJ, Hankinson SE, Hough H, Gertig DM, Garcia-Closas M. Spiegelman D. et a7. A prospective studv of NAT2 acetylation genotype, cigarette smoking. and risk of breast cancer. Carcinogenesis 1997: 18:2127-2132. Krieger N, Hiatt RA. Risk of breast cancer after benign breast diseases: variation by histologic type, degree of atypia. age of biopsy. and length of follow-up_ Arn J Epideniol 1992, .135:619-631. Lash TL. Aschengrau A. Active and passive cigarette smoking and the occurrence of breast cancer. Am I Epidemiol 1999; 149:5-12. Li D. Wang M. Dhingra K, Hittelman WN. Aromatic DNA adducts in adjacent tissues of breast cancer patients: clues to breast cancer etiology. Cancer Res 1996: 56:287-293. Lin HJ. Han C-Y, Lin BK, Hardy S- Slow acetylator mutations in the human polymorphic N-acetyltransferase gene in 798 Asians, blacks, Hispanics, and whites: application to meta- bolic epidemiology. Am J Hum Genet. 1993: 52:827-834. Lin HJ. Han C-Y, Link. Lin BK, Hardy S. Ethnic distribution of slow acetylator mutations in the polymorphic N-acetyltranv ferase (NAT2) gene- Pharmacogenetics 1994; 4:125-134. London SJConnolly JL. Schnitt SJ. Colditz GA. A prospective study of benign breast disease and the risk of breast cancer. JAMA 1992; 267:941-944. Mant D, Vessey MP, Smith MA. Yeates D. Cigarette smoking and benign breast disease. J Epidemiol Commun Heal[h 1986: 40:277. Michnovicz JJ. Hershcopf RJ, Naganuma H. Bradlow HL. Fishman J. Increased 2-hydroxylation of estradiol as a possible mechanism for the anti-estrogenic effect of cigarette smok- ing. New Engf J Med 1986: 31 S:1 305-1 309. Michnovicz JJ. Naganuma H. Hcrshcopf RJ. Bradlow HL. Fishman J. Increased urinary catechol excretion in female smokers. Steroids 1988; 52:69-83. Millikan RC. Pittman GS. Newman B, Tse C-K J. Selmin 0, Rockhill B. et al. Cigarette smoking. N-acetyltransferases 1 and 2, and breast cancer risk. Cancer Epidemiof Biomark Pren 1998: 7:371-378. Morabia A, Bernstein M. Heritier S. Khatchatrian N. Relation of breast cancer with passive and active exposure to tobacco smoke. .4m J Epidemiot 1996: 143:918-928. Nomura A, Comstock CW. Tonascia /A. Epidemiologic character- 469 isGCS nf henign breast disease. Anz J Iy0idcmiol 1977: 105:505-512. Page DL. Rogers L61'. Combined histolopathologic and cytologic criteria for the diagnosis of mamniary atypical ductal hyperplasia. Hum Paihol 1992; 23:1095-109:. Palmer JR. Rosenberg 1.. Cigarette smoking and the risk of breast cancer. Epidemiol Rev 1993: 15:145-156. pasudes H. Najjar MA. Kelsey JL, Estrogen replacement therapy and fibrocysticc breast disease. Airr J Prev Med 1987: 3:282- 286. Pfau W. O'Hare Ml, Grover Pl.. Phillips DH, Metabolic activation of the food mutagens 2-amino-3-methyhmidazo[4,5-fJquino- line (IQ) and 2-amino-3. 8-dimetltylimidazo[4.5-f)quinoxa- line (Me1Qx) to DNA binding species in human mammary epthelial cclls- Carcinogenesis 1992: 13:907-909. Philip PA. Rogers HJ. Millis RR. Rubens RD, Cartwright RA. Acetylamr status and its relationship to breast cancer and other diseases of t.he breast. Fur J Cancer Cfin Oncol 1987: 23:1701-1 706. Risch A. Wallace DM, Bathers S. Sim B- Slow N-acetylation genotype is a susceptibility factor in occupational and smoking related bladder cancer. Hum Mol Genet 1995: 4:231-236. Rohan TE Cook MG, Baron JA. Cigarette smoking and benign proliferative epithelial disorders of the breast in women: a case-control study. J Epidemiol Commun Health 1989: 43:362-368. Smith SJ, Deacon JM, Chilvers CE. Alcohol. smoking, passive smoking and caffeine in relation to breast cancer risk in young women. UK National Case-Control Study Group. Br J Cancer 1994: 70:112-119. Ssvaminathan S. Frederickson SM, Hatcher JF. Metabolic activa- tion of N-hydroxy-4-acetylaminobiphenyl by cultured hu- man breast epithelial cell line MCF 10 A. Carcinogenesis 1994: 15:611-617. Talaska C, al-Juburi AZ, Kadlubar FF. Smoking related carcino- gen-DNA adducts in biopsy samples of human urinary bladder: identification of N-(deoxyguanosin-8-yl)-4-aminobi- phenyl as a major adduct Proc Natf Acad Sci USA 199L 88:5350-5354. Thomas EJ. Edridge W. Weddell A, McGill A. McGarrigle HHG. The impact of cigarette smoking on the plasma concentra- tions of gonadotropins. ovarian steroids and androgens and upon the mctabolism of oestrogens in the human female. Hum Reprod 1993; 8:1187-1193. Vatsis KP. Martell KJ. Weber WW. Diverse point mutations in the human gene for polymorphic N-acetyltransferase. Proc Natl Acad Sci USA 1991; 88:6333-6337- Vatsis KP. Weber WW. Bell DA, Dupret JM, Evans DAP, Grant DM, et al. Nomenclature for N-acetyltransferases. Pharmaco- genetics 1995: 5:1-17. Vineis P. Bartsch H. Caparaso N, 1larrington A. KadLubar F. Landi M, et al. Genetically based N-acetyltransferase meta- bolic polymorphism and low-level environmental exposure to carcinogens. Nature 1994: 369:154-156. Wyshak G. Frisch RE. Albright NL, Albright TE. SchifT 1. Cigarette smoking and benign breast disease [letterj. N Eng! J Med 1988;319:1736-1737. Yamashita M, Wakabayashi K. Nagao M, Sato S. Yamaizumi Z. Takahashi M. ct al. Detection of 2-amino-3-methylimi- dazo[4.5-flquinoline in cigarette smoke condensate- Jpn J Cancer Res 1986: 77:419 F22.