Chapter 7

Carcinogenic Effects

7.0 Introduction

It has been estimated that 22% of all cancer deaths in women and 45% of all cancer deaths in men can be attributed to personal smoking habits (Shopland et al., 1991). Smoking is an established cause of cancers of the lung, larynx, oral cavity (including pharynx), esophagus and bladder. It is a probable cause of cancers of the kidney, pancreas, and stomach in men and women, and cervical cancer in women (IARC, 1986; U.S. DHHS, 1989). Environmental tobacco smoke (ETS) has been established as a cause of lung cancer in nonsmokers (U.S. DHHS, 1986; NRC, 1986; U.S. EPA, 1992). This document explores the role of ETS in the etiology of cancers, including lung cancer and cancers other than lung, in nonsmokers.

In this review, we first present the available data on the relationship between ETS and all cancers combined, in adults (Section 7.1.1) and in children (Section 7.1.2). Second, we discuss the evidence on the role of ETS in the etiology of specific cancer sites. Section 7.2 presents the data on ETS and lung cancer. In Section 7.3, the evidence is discussed on ETS exposure and cancer sites other than lung which are causally linked to active smoking. The evidence on ETS and cancer sites where the role of active smoking is equivocal (e.g., cancers of the breast, stomach, brain, and hematopoeitic system) is discussed in Section 7.4. That section also includes the evidence on ETS exposure and risk of childhood cancers (specific sites). Individual studies are described briefly and the results, including the point estimates of relative risks and corresponding 95% confidence intervals, are presented. Findings from the studies are evaluated, taking into account the quality of the studies with respect to their study design, sample size, assessment of exposure, adjustment for potential confounders, and consideration of sources of biases. For cancers that are causally associated with active smoking, we also compare the magnitude of the risk associated with ETS exposure versus that of active smoking.

Overall death rates for smokers are about two times higher than for nonsmokers (U.S. DHHS, 1979). Those nonsmokers who are exposed to tobacco smoke are exposed to the same toxic constituents of tobacco smoke as smokers (U.S. DHHS, 1986), although active smokers and those exposed to ETS may differ in the relative amounts of carcinogens to which they are exposed. Furthermore, the phase distributions of compounds differ between mainstream smoke and ETS. More of the constituents appear in the vapor phase (versus the particulate phases) in ETS compared to mainstream smoke, and particle sizes are smaller in ETS. Components also enter the vapor phase from the particulate phase as ETS ages. Therefore, the relative uptake and deposition of these components potentially differ between active and passive smokers (Guerin et al., 1992) (See chapter on Exposure Measurements and Prevalence). Because of these differences, it is not apparent which cancer sites may be most affected by ETS exposure. This section describes studies addressing the overall risk of cancer (all sites combined) from ETS exposure, in adults and in children.

Cancer risk in adult life may be due to an accumulation of exposures incurred transplacentally, during childhood, and during adult life. To study the potential role of ETS exposure in the etiology of various cancers in adults, most of the studies have focused on the association between adult exposure to ETS and subsequent risk (Hirayama, 1984; Sandler et al., 1989, Reynolds et al., 1987; Sandler et al., 1985a) although the role of ETS exposure during childhood and risk of adult cancers has also been investigated (Sandler et al., 1985b).

7.1.1.1 Cohort studies

Risk of all cancers in nonsmokers exposed to ETS (based on spousal smoking) was evaluated in three cohort studies.

Hirayama, 1984

In the first cohort study, the mortality of 91,540 nonsmoking wives in relation to the smoking habits of their husbands was investigated in Japan (Hirayama, 1984). Mortality of the cohort was monitored by review of death certificates and the annual census of residents. After 16 years of follow-up, there were a total of 2705 cancer deaths (all sites) among the nonsmoking women. The relative risks (RRs) were 1.00, 1.12 (95% CI=1.03-1.21), and 1.23 (95% CI=1.12-1.35) for women whose husbands were nonsmokers, ex-smokers or smokers of 1-19 cigarettes per day, and smokers of 20 or more cigarettes per day, respectively, when adjustment was made for husband's age and occupation. In this population, the increased risk for all cancers combined was due mainly to the increased risk observed for cancers of the lung, nasal sinus, and brain. These respectively accounted for 7%, 1%, and 1% of the tumors in this population. Stomach cancer, representing 31% of the cancers in this population, was not associated with passive smoking (Hirayama, 1984). There was a small increased risk of cervical cancer in passive smokers (see Section 7.2.2) (Table 7.1).

Using a cohort surveyed in 1963 in Western Maryland, Sandler et al. (1989) evaluated the all-cancer mortality in nonsmokers who lived with smokers. A total of 22,973 Caucasian men and 25,369 Caucasian women were enrolled; 4,162 men and 14,873 women were lifetime nonsmokers. In 1975, death records were reviewed to evaluate the risk of mortality, and specific causes of mortality, in passive smokers compared to nonsmokers not exposed to ETS. In brief, a score ranging from 0 to 12 was assigned to each adult in the household based on his/her smoking history. A total household smoking score was then calculated by summing the smoking contribution scores of all persons living in that household. Each individual's household ETS exposure was calculated by subtracting his or her own contribution from the total household score. Among nonsmokers, 1,248 men (30.0%) and 9,551 women (64.2%) were exposed to household tobacco smoke and were considered to be passive smokers. Exposure to ETS did not increase the risk for all cancers combined in nonsmoking men (RR=1.01, 95% CI=0.66-1.53) and nonsmoking women (RR=1.00, 95% CI=0.82-1.21) after adjusting for age, marital status, education, and housing quality. When the analysis was conducted separately for tumors related to smoking and tumors not related to smoking, exposure to ETS was associated with a small increased risk for smoking-related tumors in women (RR=1.45, 95% CI=0.88, 2.40), but not in men (RR=0.96, 95% CI=0.66-1.53). In men and women, there was no association between ETS exposure and risk of nonsmoking-related tumors (Table 7.1).

Reynolds et al. (1987) reported results from a small cohort of 2413 married women (46% had never smoked) who participated in a population-based survey in Alameda County, California in 1965. Smoking history was independently ascertained for each spouse. Based on 71 cancers diagnosed among the 1111 nonsmoking women during the 17 years of follow-up, nonsmoking women whose husbands smoked showed a RR of 1.68 (90% CI = 1.1-2.5) for all cancers combined compared to women whose husbands did not smoke. The authors also reported a 7-fold increased risk (90% CI=1.1-47.0) of smoking-related cancers in relation to husband's smoking (Table 7.1), but this was based on 4 cases only (smoking-related cancers included cancers of the lung, mouth, esophagus, bladder, pancreas, liver, kidney, and uterine cervix); the specific sites of the 4 cases were not presented.

7.1.1.2 Case-control studies

Overall cancer risk in relation to ETS exposure from spouses and parents was evaluated in a case-control study conducted by Sandler et al. (1985a). This study included all cancers (excluding skin cancers) diagnosed between ages 15 to 59, during July 1979 through March 1981, from the hospital-based tumor registry affiliated with the University of North Carolina. Of the 740 eligible cancer cases, 518 completed a mailed questionnaire which included information on ETS exposure during childhood and adult life. For 360 of the 518 cases, a friend/acquaintance of the same race, gender, and within five years of age of the case served as a control in the study. The remaining controls were identified by systematic telephone sampling using the telephone numbers of the cases as a starting point. Passive smoke exposure during childhood was based on whether their natural parents ever smoked, smoked before the subject's birth, smoked in the house for most of the years before the subject was 10 years old, and whether mothers smoked while pregnant with the index subject. Passive smoke exposure during adult life was based on the number of years of marriage during which a spouse smoked at least 1 cigarette per day for as long as 6 months. The average number of cigarettes smoked by spouses was also obtained. Among the 518 cases and controls, 231 cases and 235 controls were lifetime nonsmokers.

Among lifetime nonsmokers, there was a significant 2-fold increased risk (95% CI=1.4, 3.0) associated with spouses' smoking after adjustment for gender, race, and age. When the effect of ETS exposure was examined by age group, gender, and race, the effect was more apparent for subjects aged 40-49 (adjusted RR=2.0, 95% CI=1.4-2.9), females (adjusted RR=2.0, 95% CI=1.3-2.9), and non-whites (adjusted RR = 2.0, 95% CI = 1.4 - 3.0). However, no dose-response relationship was observed between risk and the number of years married to a smoker or number of cigarettes husbands smoked per day (data not presented). The role of ETS exposure was also investigated by site of tumor. The increased risk was not limited to lung cancer and other smoking-related tumors, such as cervical cancer. Increased risks were also observed for breast and endocrine gland cancers, tumors not causally associated with active smoking.

In a second report of the same adult study population, Sandler et al., (1985b) evaluated the association between ETS exposure from parents and risk of all cancers. Mother's and father's smoking habits were available on 438 cases and 470 controls; 197 cases and 223 controls were lifetime nonsmokers. Maternal and paternal smoking was each associated with a nonsignificant 20 percent increased risk for all cancers among nonsmokers. The effect of maternal and paternal smoking was evaluated for `smoking-related' and `non-smoking related' cancers. `Smoking-related' cancers included tumors of the oral cavity and pharynx, esophagus, pancreas, respiratory and intrathoracic organs, urinary tract and cervix, and accounted for some 25% of tumors in nonsmokers. For ‘smoking-related' tumors, the RR was 0.76 (95% CI=0.25, 2.30) for maternal smoking and 1.68 (95% CI=0.86, 3.29) for paternal smoking. For cancers not related to smoking, the RR was 1.24 (95% CI=0.65, 2.36) for maternal smoking and 1.13 (95% CI=0.73, 1.75) for paternal smoking.

7.1.1.3 Summary

In summary, there is limited evidence from two cohort studies (Hirayama, 1984; Reynolds et al., 1987) and one case-control study (Sandler et al., 1985a) that exposure to spouses' smoking may increase overall risk of cancer in nonsmoking women. In one study, the increase is explained primarily by an elevated risk observed for lung cancer (Hirayama, 1984). However, in two studies, elevated risks were observed for sites not typically related to active smoking as well as sites related to smoking (Reynolds et al., 1987; Sandler et al., 1985a). In the study by Reynolds et al. (1987), the strong association between husbands' smoking and smoking-related tumors was based on very few cases, accounting for only 6% of all cancers. In the study by Sandler et al. (1985a), increased risks were observed for both smoking-related (lung, cervix), and non-smoking-related sites (breast and endocrine gland) after adjustment for age and education. Although the results on nonsmoking-related cancers are intriguing, they are difficult to interpret given that known risk factors for the specific cancers under study were not adjusted for (Sandler et al., 1985a). Possible effects of potential confounders are a concern and in further studies should be more carefully researched. For example, sexual activity is a risk factor for cervical cancer and exposure to ETS may be associated with sexual activity. Alcohol intake is a risk factor for breast cancer and exposure to ETS may be positively associated with alcohol use.

Exposure to ETS has been investigated as a risk factor for all childhood cancers combined and for specific childhood tumors (see Sections 7.3.3 to 7.3.6). Exposure to ETS may occur during the prenatal or postnatal period. Prenatally, the fetus may be exposed to tobacco smoke constituents when the mother smokes during pregnancy (i.e., transplacental effects) or if the mother is exposed to someone else's smoking, most likely the father's smoking. Postnatally, the child may be exposed to ETS directly by inhalation. The main sources of postnatal ETS exposure to a child whose parents both smoke is likely to be from the mother, and to a lesser extent the father.

In this chapter, mothers' smoking during pregnancy is considered to be a surrogate measure of mothers' smoking postnatally (see below). However, since studies on childhood cancers included subjects who were diagnosed with cancer up to age 24, it is reasonable that tobacco smoke exposure both in utero and postnatally would be important. Thus, study findings require cautious interpretation.

The extent of information on passive smoke exposure varied in the different studies. Two case-control studies conducted in the 1950's asked about mothers' or fathers' smoking habits at the time of interview or study enrollment. In one study, this pertained to smoking habits of parents at the time of interview which was after the death of the subject under study (Stewart et al., 1958). The other study, focused on the mother's smoking habits when study subjects were enrolled (Manning and Carroll, 1957). In more recent studies, mothers' smoking habits during pregnancy were available (Neutel and Buck, 1971; Stjernfeldt et al., 1986; Pershagen et al., 1992; Severson et al., 1993). Several studies offered more detailed information by including mothers' smoking habits 1-2 years before and during the pregnancy (Gold et al., 1979; Van Steensel-Moll et al., 1985; John et al., 1991; Gold et al., 1993). Mothers' smoking during pregnancy represents transplacental exposure to tobacco smoke constituents and may also be used as a proxy variable of postnatal ETS exposure of the child. There are data to support the assumption that mothers' smoking habits during pregnancy represent an unbiased estimate of their smoking habits after pregnancy. In a study of childhood cancers and maternal smoking (Stjernfeldt et al., 1986), comparison of mothers' smoking habits 5 years before, during, and after pregnancy showed that a similar percentage (8%) of cases' and controls' mothers reported they smoked after pregnancy when they had not smoked during pregnancy. In a study of childhood brain tumors (Gold et al., 1993), comparable percentages of mothers of cases (72%) and of population controls (73%) who had ever smoked reported they were smoking during the birth year of the child. However, some women may quit during pregnancy and resume afterwards so there is potential misclassification when smoking status is based only on smoking habits during pregnancy.

Other studies on childhood cancers obtained information on both mothers' and fathers' smoking habits during the index pregnancy (Preston-Martin et al., 1982; McKinney and Stiller, 1986; Buckley et al., 1986; Howe et al., 1989; John et al., 1991; Gold et al., 1993; Kuijten et al., 1990; McCredie et al., 1994). Children whose non-smoking mothers were exposed to spouses' smoking were thus considered exposed to ETS prenatally. In some studies, the effect of fathers' smoking was evaluated among children of non-smoking mothers (John et al., 1991; McCredie et al., 1994; Gold et al., 1993). None of the studies collected information on father's smoking postnatally. However, on the basis of the above mentioned data which showed that mother's smoking during pregnancy is an unbiased estimate of her smoking postnatally (Stjernfeldt et al., 1986) or ever smoking (Gold et al., 1993), we assume that father's smoking during pregnancy is also an unbiased proxy for father's smoking postnatally.

7.1.2.1 Biomarkers studies of exposure to tobacco smoke constituents in utero and postnatally

The effects of transplacental exposure to tobacco smoke constituents due to maternal active smoking during pregnancy are difficult to distinguish from those of postnatal ETS exposure. Recent studies investigating the levels of three different biomarkers of tobacco smoke exposure in mothers and their offspring have demonstrated that the fetus (Coghlin et al., 1991; Hammond et al., 1993), the newborn (Eliopoulos et al., 1994) and the young child (Crawford et al., 1994) are exposed to considerable amounts of tobacco products.

In one of the studies (Eliopoulos et al., 1994), mothers were identified 1 to 3 days after delivery and 5 to 7 hair shafts were obtained near the skull from both the mothers and their newborns for determination of nicotine and cotinine levels (Table 7.1b). Although previous studies typically measured cotinine and nicotine levels in saliva, serum, or urine, levels measured in hair samples provide more long-term assessment of ETS exposure. Nicotine and cotinine levels were highest in mothers who were active smokers, intermediate in nonsmokers who were passive smokers, and lowest in nonsmokers not exposed to ETS. The respective mean levels were 19.2, 3.2, and 1.2 for nicotine (ng/mg) and 6.3, 0.9, and 0.3. for cotinine (ng/mg). Newborns of smokers showed significantly higher mean levels of nicotine (2.4 ng/mg) than newborns of passive smokers (0.28 ng/mg) or nonsmokers (0.4 ng/mg). Nicotine levels in newborns of passive smokers were not higher than those of nonsmokers but the difference in levels was not statistically significant. On the other hand, mean levels of cotinine were highest in newborns of smokers (2.8 ng/mg), intermediate in passive smokers (0.6 ng/mg) and lowest in nonsmokers (0.26 ng/mg). The cotinine levels in newborns of passive smokers were significantly higher than those of nonsmokers and significantly lower than those of smokers. The authors explained that nicotine may be a less sensitive marker than cotinine because of its shorter half-life (1-3 hours for nicotine compared to 10-14 hours for cotinine).

In a study conducted by Coghlin et al. (1991), maternal-fetal exchange of a potent tobacco-related human carcinogen, 4-aminobiphenyl (4-ABP), was studied in smoking (n=14) and nonsmoking (n=38) pregnant women. N-hydroxy-4-ABP, the active metabolite of 4-ABP, forms chemical adducts with hemoglobin. Levels of 4-ABP hemoglobin adducts were detected in all maternal-fetal paired blood samples. The mean levels of such adducts were 183 (pg/g of hemoglobin) in smoking women, 92 in fetal blood samples from smokers, 22 in nonsmoking women, and 17 in fetal blood samples from nonsmokers. In a related study conducted by the same investigators (Hammond et al., 1993), the relationship between levels of 4-ABP-hemoglobin adducts and exposure to ETS in nonsmoking women (based on nicotine levels measured by passive monitors) was investigated. The median level of 4-ABP adduct was 26 pg/g among nonsmoking women in the highest ETS exposure category (³ 2 m g/m3 weekly average nicotine) compared to median levels of 15 pg/g among those with the lowest ETS exposure (<0.5 m g/m3 weekly average nicotine). The levels of 4-ABP hemoglobin adducts in nonsmoking women were 12% of those in smokers whereas levels in fetuses of nonsmoking women were about 9% of those of smoking women. These two studies provided evidence that 4-ABP crosses the human placenta and binds to fetal hemoglobin in both nonsmoking and smoking mothers and that among nonsmoking women, the levels of 4-ABP adducts increased significantly with increasing levels of ETS exposure (Hammond et al., 1993).

In the third study, Crawford et al. (1994) evaluated levels of serum cotinine and polycyclic aromatic hydrocarbon (PAH)-albumin adducts in Hispanic and African-American preschool children and their mothers. In this study, mean serum cotinine levels were highest in mothers who smoked (170 ng/ml), intermediate in nonsmoking mothers exposed to passive smokers in the household (1.64 ng/ml), and lowest in nonsmoking mothers not exposed to ETS in the household (0.96 ng/ml). A similar gradient in serum cotinine was observed in preschool children whose mothers were smokers (4.14 ng/ml), passive smokers (0.87 ng/ml), and nonsmokers not exposed to household ETS (0.25 ng/ml). Levels of PAH-albumin adducts (fmol/ug) followed the same pattern in mothers who were smokers, passive smokers and nonsmokers; the respective levels were 0.80, 0.49, and 0.31. In preschool children of smokers, passive smokers, and nonsmokers not exposed to ETS, the corresponding levels of PAH-albumin adducts were 0.35, 0.18, and 0.15. Comparisons between the three groups of mothers and between the three groups of preschool children show that there were statistically significant differences in levels of cotinine and PAH-albumin adducts, with those in smokers (or their children) higher than those in passive smokers and nonsmokers not exposed to ETS (or their children). Although levels in passive smokers (or their children) were also higher than those in nonsmokers not exposed to ETS (or their children), the differences were not statistically significant. Levels of cotinine and PAH-adducts in children whose mothers were passive smokers (i.e., exposed to household ETS) were lower (levels were about one-third to one-half) than those of their mothers who were living in the same ETS-exposed households, presumably because mothers had more opportunities to be exposed to ETS outside the home than did their preschool children.

In this study, young children exposed to ETS via their mothers' smoking showed increases in cotinine and PAH-albumin adducts. These results suggest that exposed children can take up and metabolically activate respiratory carcinogens. Children with nonsmoking mothers who were exposed to ETS from other household members also showed increases in levels of cotinine and PAH-albumin adducts, although the increases were smaller.

7.1.2.2 Cohort Studies

Two prospective studies and a case-cohort study investigated the effect of maternal smoking during pregnancy and risk of cancer in children (Table 7.3).

A study by Neutel and Buck (1971) was based on 89,302 births registered in Canada and the United Kingdom. The cohort included all births registered in 10 Canadian hospitals between 1958 and 1961, as well as those registered in all hospitals in England and Wales during a one-week period. Smoking habits of mothers during pregnancy were recorded before or just after the birth of the child. For 74% of the cohort (n=66,456), mothers were classified as nonsmokers, smokers of less than one pack per day, or smokers of one or more packs per day. In the remainder of the cohort, nonsmoking mothers and those smoking less than one pack per day could not be distinguished and thus are excluded from this discussion. A total of 65 cancer deaths (22 leukemias, 20 nervous system tumors, and 23 other sites) occurred before age 10 among the 66,456 births. There was a small increased risk for all cancers combined (RR=1.31, 95% CI=0.8, 2.2) among children whose mothers smoked compared to children whose mothers did not smoke. There were few cases in the heavy smoking category and a consistent dose trend of increasing risk with increasing amounts smoked by mothers during pregnancy was not observed.

A second cohort study was conducted by Pershagen et al. (1992) who utilized data from the Swedish Medical Birth Registry and the Swedish Cancer Registry. Cancer incidence in a cohort of 497,051 children born between 1982-1987 was determined and compared by maternal smoking at 2-3 months of pregnancy (none, <10 cigarettes/day, or > 10 cigarettes/day). Relative risks were adjusted for potential confounders which included maternal age, birth order, year and county of birth of index subject. There were a total of 327 cancers for which maternal smoking habits were known: 198 solid tumors and 129 tumors of the lymphatic and hematopoietic system. There was no association between maternal smoking and risk of all cancers combined (adjusted RR=0.99, 95% CI=0.78-1.27). The lack of an association persisted when the analysis was conducted separately for solid tumors combined (adjusted RR=0.96, 95% CI=0.70-1.32), and for lymphatic and haematopoietic tumors combined (adjusted RR=1.04, 95% CI=0.71-1.52).

The study by Pershagen et al., (1992) has several strengths but a major limitation. The compilation of the cohort of births was nearly complete (99%). Of the 422 childhood cancer cases identified in the Swedish Cancer Registry during this time period, 408 could be linked to a subject in the birth cohort (we assumed that 81 of 408 cases were excluded from the analysis because data on maternal smoking habits were missing). Data on mothers' smoking habits at 2-3 months of pregnancy were available on over 90% of children born between 1983 to 1987 and for about 50 percent of children born in 1982. The lower figure in 1982 was due mainly to logistical problems during this first year when the birth registry started to collect information on smoking. Results remained unchanged when births in 1982 were excluded from the analysis. The percent of mothers who smoked in this study was also similar to that reported in other Swedish studies, so that underreporting of smoking during pregnancy cannot explain the lack of an association. The main limitation is that the maximum follow-up was to 5 years of age and thus an effect of maternal smoking on cancers occurring at older ages was not assessed; there were small numbers of cancers diagnosed among the 4-5 year olds.

A case-cohort study was conducted by Golding et al. (1990), who collected information prospectively on 16,193 infants delivered over a one week period in 1970 in the United Kingdom. These children were followed up at ages 5 and 10; 80% and 94% respectively were successfully contacted. By 1980, 33 children had developed cancer (9 leukemia, 5 lymphoma, 8 brain, 5 Wilm's tumor, 6 other). For each cancer case, 3 controls were selected and matched to cases on factors including maternal age at birth of index subject, parity and social class. Significantly more mothers of cases had smoked 5 cigarettes or more per day throughout pregnancy compared to the controls (RR=2.47, 95% CI = 1.2, 5.1). Maternal smoking remained statistically significant in logistic regression analysis when other risk factors (e.g., social class, x-ray in pregnancies, use of various medications) were controlled for.

7.1.2.3 Case-control studies

One of the first case-control studies to examine the role of parental smoking and risk of childhood cancers was a hospital-based study conducted in the U.S. (Manning and Carroll, 1957). Smoking habits of mothers (at time of study enrollment) of children with cancers (188 leukemias, 42 lymphomas, and 93 other cancers) were compared to mothers of children with orthopedic diseases (n=50). There was no difference in the percent of mothers of children with cancer who smoked 10 or more cigarettes per day (37.4%) compared to mothers of controls (38.0%). A second study was conducted by Stewart et al. (1958) who included as cases all children in England and Wales who had died of leukemia or other cancer before their 10th birthday between 1953 and 1955. Controls were individually matched to cases on gender, age (plus or minus 6 months of the birth date of the cases) and locality of residence. A total of 1416 case/control pairs were available for analysis. Fathers and mothers of the index subjects were classified as heavy, moderate, light, or non-smokers. The smoking habits of fathers of children with cancer (82.9% smoked at least one cigarette or pipe per day) were similar to those of fathers of control children (80.9% smoked). There was a small excess of mothers of cases who smoked (47.8%) compared to mothers of controls (43.8%) (OR=1.09, p=0.04) but this was not adjusted for potential confounding factors. The authors cautioned that since parents were interviewed after the death of the index patients, their smoking habits may be affected by bereavement.

Results from 5 case-control studies conducted since the 1980's offer better information on smoking habits of parents during pregnancy (Table 7.3).

Stjernfeldt et al. (1986a; 1986b; 1992) conducted a population-based, nationwide case-control study of childhood cancer in Sweden. A total of 305 children, aged 16 or younger, diagnosed with cancer during 1978 and 1981 were identified by the Swedish Child Leukemia Group. Cases were compared to 340 control children with insulin-dependent diabetes mellitus. Families of cases and controls completed a self-administered questionnaire with an overall participation rate of about 95% in both groups. Controls were not individually- or frequency-matched to cases on age or gender, but these variables were controlled for in the analysis. Information on smoking habits of mothers was obtained on 92 percent of cases and controls for the 5-year period before pregnancy, during pregnancy, and postnatally to onset of disease in the index subject.

There was some suggestion of an increased risk for all cancers combined in relation to mother's smoking during pregnancy. Compared to children whose mothers were nonsmokers, children whose mothers smoked 1-9, and 10+ cigarettes per day showed RRs of 1.07 (95% CI=0.63-1.80) and 1.56 (95% CI=1.05-2.33) respectively. The increase in risk was not observed for solid tumors but was restricted to tumors of the reticuloendothelial system, primarily acute lymphoblastic leukemias. The authors did not present results separately for mothers' smoking after birth of the index subject, but suggested that since mothers who smoked during pregnancy generally smoked after the child was born, it would be difficult to separate the effect of in utero exposure to tobacco smoke constituents versus postnatal ETS exposure.

Despite concerns raised regarding the choice of controls and possible selective recall bias among cases (McKinney and Stiller, 1986; Buckley et al., 1986; Dahlquist and Wall, 1986; Li, 1986; Cunningham, 1986), none of these biases appear to explain the study's findings. It can be argued that mothers of diabetic children would recall more similarly to mothers of children with cancer if there is any recall bias associated with having a disease. Smoking habits of mothers of diabetic children were representative of the general population since their smoking prevalences were comparable to those of Swedish women surveyed in studies conducted during the same time period. Moreover, differences between cases and controls in ages at diagnosis, geographic location, and socio-economic status could not explain the apparent findings (Stjernfeldt et al., 1986b). The increased risk associated with maternal smoking was observed after adjustment for factors including maternal age, birth order of index subject, and parental occupation.

In response to the findings of Stjernfeldt et al. (1986), McKinney and Stiller (1986) published a letter to the editor and a more detailed paper (McKinney et al., 1987) presenting data collected for the Inter-Regional Epidemiology Study of Childhood Cancer (IRESCC), a collaborative study conducted in three health regions in the United Kingdom (Yorkshire, West Midlands, and North West) between 1980 and 1983. Study subjects included 555 children (< age 15) diagnosed with childhood cancer. Two healthy, age- and sex-matched control children were identified for each case using the general practitioner lists and admissions to hospital for minor conditions. Parents of cases and controls were asked identical questions regarding the antenatal period of the index subject (e.g., illness, use of medications, complications, smoking and drinking habits) (McKinney et al., 1987). Maternal smoking habits during pregnancy were not associated with risk for all cancers combined; the RRs were 1.0, 1.12 (95% CI = 0.85, 1.47) and 0.84 (95% CI = 0.65, 1.09) respectively, for mothers smoking 0, 1-10, 11 + cigarettes/day. Leukemias and lymphomas, which accounted for 44% of the childhood cancers in this population, were not associated with maternal smoking. However, maternal smoking was associated with nonsignificant increased risks for soft tissue sarcomas and bone tumors (see Section 7.3.6: Bone and Soft-tissue Sarcomas).

Also in response to Stjernfeldt's findings, Buckley et al. (1986) investigated the role of maternal smoking during pregnancy and the risk of childhood cancer using data gathered by the US Children's Cancer Study Group. Since 1983, parents of 1814 children have completed a questionnaire which included smoking histories of the mother and father before and during the pregnancy of the index subject. Controls were drawn at random from approximately the same geographic regions of cases in the US and Canada. There was no association between maternal smoking during pregnancy and risk of all cancers combined; the RRs were 1.31, and 0.97 respectively, for mothers smoking 1-9, and 10+ cigarettes/day during pregnancy compared to nonsmokers. Acute lymphoblastic leukemia, representing 41% of cancers in this study, was not related to mother's smoking. Paternal smoking during the index pregnancy was also not associated with all childhood cancers combined (data were not presented). Adjustment for potential confounders (e.g., birth year of the child, maternal age, illnesses during the pregnancy, and socioeconomic factors) did not alter the results.

John et al. (1991) investigated the role of parental smoking before and during pregnancy and the risk of childhood cancer in a population-based case-control study conducted in Colorado. The study included incident childhood cancers, diagnosed between 1976 to 1983 among children 14 years old or younger. Controls were selected by random digit dialing and were individually matched to cases on age (± 3 years), sex, and telephone exchange area. Of the 356 eligible cases, 252 (response rate of 70.8%) participated in the study compared to 222 controls (response rate of 62.8%). Structured interviews were administered to parents of index subjects and included questions on smoking habits of mothers, fathers, and other household members during the index pregnancy. In addition, questions regarding the mother's cigarette smoking habits at three months prior to the index's conception and during each trimester of the pregnancy were asked. Questions on father's smoking included use of cigarettes, cigars and pipes. Information on other smokers in the household was derived based on questions regarding the number of regular smokers at each residence from conception to the time of the child's diagnosis. The definition of nonexposed in this study is not exposed to smoking by either parent or other household members from the period starting one year before birth through the time of diagnosis. Data on the number of cases and controls who were exposed to other household members only (but not to parents' smoking) were not presented.

For all cancers combined, there was a small increased risk associated with exposure to mothers' or fathers' smoking. The RR for all cancers was 1.3 (95% CI=0.8-2.0), 1.5 (95% CI=1.0-2.5), and 1.4 (95% CI=0.9-2.4), respectively, in relation to mothers who smoked during the 3 months prior to conception, the first trimester, and all three trimesters of the pregnancy. The ORs for all cancers combined was 1.3 (95% CI=0.9-2.0) in relation to any tobacco use by the father. The ORs for all cancers combined in association with mothers' smoking in the absence of father's smoking, fathers' smoking in the absence of others' smoking, and the combined effect of mothers' and fathers' smoking were 1.7 (95% CI=0.7-4.3), 1.4 (95% CI=0.9-2.3), and 1.5 (95% CI=0.9-2.6), respectively. The data suggest an increasing trend in risk with increasing amounts smoked by mothers, but not by fathers. The positive association between ETS exposure and risk of all cancers is largely due to its effect on risk for acute lymphoblastic leukemia, lymphoma and brain tumors. Father's education was a potential confounder in this study. The OR for all childhood cancers in relation to fathers’ and mothers’ smoking was 1.5 (95% CI = 0.9 - 2.6); this OR was reduced to 1.2 (95% CI = 0.7 - 2.1) when father's education was accounted for in the analysis.

7.1.2.4 Summary

While in some studies increased risks overall in childhood cancers were observed, in others no such increases were seen. There are several limitations in both the studies finding an association and those finding no association between ETS exposure and risk of childhood cancers. The cohort study of Pershagen et al. (1992) is limited in that it can only examine the effect of ETS exposure on tumors diagnosed up to 5 years of age whereas all the other studies included cancers up to 10 or 16 years of age. Causes of childhood cancers in very young children may differ from those of older children. The two large case-control studies which found no association with maternal smoking were collaborative studies of childhood cancers conducted in the United Kingdom (McKinney and Stiller, 1986) and the U.S. (Buckley et al., 1986). Selection bias of cases cannot be ruled out in these studies. Childhood cancer patients admitted to academic institutions were enrolled in these studies and these may be unrepresentative of all childhood cancers in the population (e.g., higher social class). The denominator of childhood cancers was not presented and thus participation rates could not be calculated. Because of the association between social class/education and smoking habits, selection bias associated with social class/education cannot be precluded. Prevalence of smoking habits of mothers/fathers were not presented in these two studies. On the other hand, the strongest positive finding reported in the case-cohort study by Golding et al. (1990) was based on a small number of cases and classification of mother's smoking as less than 5 versus greater than 5 cigarettes/day. The choice of the less than 5 cigarettes/day as the baseline category was not explained and it is unclear whether this cut-off was an a priori decision. Presenting the results using nonsmoking mothers as the baseline group would have been a useful comparison to other studies. The results by Stjernfeldt et al. (1986) have also been questioned because of the choice of controls (children with diabetes). Finally, there is some suggestion that inadequate adjustment for paternal education (as a surrogate for social class) may have produced an association between parental smoking and risk of childhood cancer that is artificially strengthened (John et al., 1991).

In summary, the evidence for a role of parental smoking and childhood cancers is inconclusive. One (Neutel and Buck, 1971) of two cohort studies reported an elevated risk which is not statistically significant (OR=1.3, 95% CI=0.8-2.2). Two (Stjernfeldt et al., 1986; Golding et al., 1990) of five recent (i.e., conducted in the 1980s) case-control studies reported significant associations between mother's smoking during pregnancy and risk of childhood cancers. A third case-control study (John et al., 1991) which reported elevated risks that were not statistically significant was the only study in which fathers' smoking during pregnancy in the absence of mothers' smoking was evaluated; these investigators found a statistically nonsignificant increased risk associated with fathers' smoking alone (OR=1.4, 95% CI=0.9-2.3). The positive findings are due largely to the significant association between maternal smoking and acute lymphoblastic leukemia in these studies. No other cancer site appeared to be significantly affected by maternal or paternal smoking.

Active smoking is firmly established as a causal factor for lung cancer. The Surgeon General (U.S. DHHS, 1986), National Research Council (NRC, 1986), and U.S. EPA (1992) have reviewed epidemiologic studies investigating the role of ETS exposure as a cause of lung cancer in nonsmokers. Our review focuses on studies published since the latest review: three large U.S. population-based case-control studies (Stockwell et al., 1992; Brownson et al., 1992; Fontham et al., 1991 and 1994) and a fourth, considerably smaller, hospital-based case-control study (Kabat et al., 1995).

In 1981, the first epidemiological studies of ETS exposure and lung cancer were published (Hirayama et al., 1981; Trichopoulos et al., 1981). These studies found that nonsmokers married to smokers showed a significantly higher risk of lung cancer than nonsmokers married to nonsmokers. Some 30 epidemiological studies have since been published. Most of the individual studies found a small increased risk, and a few found statistically significant results; however, all the studies published in the 1980s had small sample sizes which lacked statistical power to detect small associations. The Surgeon General (U.S. DHHS, 1986), NRC (1986), and U.S. EPA (1992) conducted comprehensive reviews of the epidemiological literature and concluded that ETS exposure was causally associated with lung cancer. Their conclusions were based on the total weight of evidence, and not on any individual study.

The U.S. EPA (1992) report reviewed a total of 30 epidemiologic studies (four prospective follow-up and 26 case-control studies) from eight countries. All the studies examined the risk of lung cancer in nonsmokers in relation to spousal smoking habits. Each study was examined in detail and then the studies were examined collectively. Because none of the studies were exactly alike, and the individual studies had different methodologic strengths and weaknesses, the U.S. EPA report ranked the studies in four tiers and gave special consideration to the 15 studies in the two highest tiers. The U.S. EPA report concluded that ETS is responsible for approximately 3000 lung cancer deaths per year in U.S. nonsmokers.

In order to gain a more accurate estimate of the association between ETS exposure and lung cancer, a meta-analysis approach has been used to pool results of comparable studies. Numerous meta-analyses have been published on this subject (U.S. DHHS, 1986; NRC, 1986; U.S. EPA, 1992; Fleiss and Gross, 1991; Arundel et al., 1987; Kilpatrick, 1992; Pershagen, 1992; Vainio and Partensen, 1989; Repace and Lowry, 1990; Spizer et al., 1990; Wells et al., 1991; Wells, 1993). A widely disseminated and reviewed meta-analysis was conducted by the U.S. EPA (U.S. EPA, 1992; Farland et al., 994; Jinot and Bayard, 1994). Despite careful considerations of many methodologic issues of concern in the meta-analysis of ETS exposure and lung cancer (e.g., measurement of ETS exposure, misclassification bias of nonsmoker status and disease status, adjustment for potential confounders), the U.S. EPA report was criticized (LeVois and Layard, 1994; Gori, 1994a & b). Some of the concerns centered around issues that were specific to the study of ETS exposure and lung cancer, including misclassification bias of smokers as nonsmokers and the extent of such misclassification. On the other hand, other issues were generic to meta-analysis techniques, and they include possible publication bias of positive studies and the difficulty in obtaining adjusted risk estimates (Gori, 1994a & b) for meta-analysis. The issue of publication bias has been reviewed in detail by Bero et al. (1994), who concluded that there is no publication bias against statistically nonsignificant results on ETS in the peer-reviewed literature.

The U.S. EPA’s (1992) reporting of 90% confidence intervals has gained much attention and is worth addressing here. The U.S. EPA report uses a one-tailed test of statistical significance (with p = 0.05) and reports the corresponding 90% confidence intervals, consistent with the one-tailed test. Use of a one-tailed statistical test could be considered to increase the probability of accepting an association (for an individual study) that occurs by chance. A one-tailed test is a standard statistical methodology used when there is prior evidence that the effect of an agent is likely to be in one specific direction. In this case, the Suregon General (U.S. DHHS, 1986), NRC (1986), and an International Agency for Research on Cancer work group (IARC, 1986) all previously concluded that ETS exposure increased lung cancer risk. The established causal association between active smoking and lung cancer, and the chemical similarity between mainstream smoke and ETS, were considered by U.S. EPA (1992) to provide prior evidence that any effect of ETS on lung cancer would be likely to be positive (i.e., to increase the risk); thus, the one-tailed significance test was the appropriate method for evaluating the hypothesis of an effect of ETS on lung cancer risk (U.S. EPA, 1994). Had U.S. EPA used a two-tailed statistical significance test (with corresponding 95% confidence intervals) instead of a one-tailed test (with 90% confidence intervals), the overall conclusions regarding causality and degree of risk would have been the same (U.S. EPA, 1994).

Three large U.S. population-based case-control studies have been published since 1991 which confirm and extend the results of the pooled U.S. studies presented in the U.S. EPA report. These studies were conducted in Florida (Stockwell et al., 1992), Missouri (Brownson et al., 1992), and in five geographic areas of the U.S. (New Orleans, Louisiana; Atlanta, Georgia; Houston, Texas; Los Angeles County, California; and San Francisco Bay Area, California) (referred to as the U.S. multicenter study) (Fontham et al., 1991 and 1994). Preliminary findings from the U.S. multicenter study (Fontham et al., 1991) were included in the U.S. EPA (1992) report. A fourth study, which is a considerably smaller, hospital-based case-control study, was published in 1995 (Kabat et al., 1995).

We will review these four U.S. case-control studies, describing their respective study designs and the main findings. In our evaluation of the methodologic issues related to the study of ETS exposure, we will focus on the sources of cases and controls, the methods used to obtain information on the exposures of interest, the verification of the exposures of interest and of the diagnosis of lung cancer, and the consideration of potential confounding variables in the analysis of ETS exposure.

To minimize confusion, we cited the exact ORs and confidence intervals that were reported in the original papers. This means that some numbers are reported to one decimal place whereas some to two decimal places. Odds ratios that we calculated are prefaced by 'we calculate' and these estimates are 'crude odds ratios'. In some instances, the numbers of cases and controls (presented in the tables) by various intensity of ETS exposure (i.e., pack-years, years of exposure) did not add up to the total numbers of subjects included in the individual studies and we assume these differences in numbers are due to missing information on specific parameters of intensity of ETS exposure or on the covariates included in the adjustments (the variables that were adjusted for in the different analyses are described as footnotes in the various tables). The measures of intensity of exposure were generally in terms of years (or smoke-years) or pack-years of exposure, number of cigarettes (or tobacco products) smoked per day, or the number of smokers in the household.

Stockwell et al. (1992) conducted a population-based case-control study of women in 28 counties in central Florida (Table 7.4). Eligible cases included women diagnosed with a histologically confirmed primary lung cancer between April 1, 1987, and February 28, 1991, and were identified through the Florida Statewide Cancer Registry and the tumor registries of area hospitals. Age criteria for the study subjects was not specified. Population controls were selected by random-digit dialing; it is unclear whether cases and controls were frequency-matched on any criteria. All cases and control subjects were lifetime nonsmokers, defined as having smoked for a total of less than 6 months or less than 100 cigarettes in their lifetime. The nonsmoking status of the study subjects was verified by checking 1) medical records and with physicians' offices (for cases) and 2) at the time the subjects were contacted to set up the interview and at the beginning of the interview (for cases and controls). The response rate for lung cancer cases was 83%; it was not specified for controls.

A combination of telephone (51% for cases; 46% for controls) and in-person (41% for cases; 54% for controls) interviews and mailed questionnaires (8% for cases; 0.3% for controls) were used to obtain information from study subjects. Interviews of surrogate respondents (primarily husbands and children) were necessary for 66.7% of the case patients who were too ill to be interviewed or were deceased. Information was obtained on a total of 210 lung cancer patients and 301 controls.

Subjects were asked about their exposure to ETS from husbands, mothers, fathers, siblings, and other household members and at the workplace. Compared to unexposed individuals who had no household ETS exposure, women who were exposed to husbands’ smoking had ORs of 1.6 (95% CI=0.8,3.0) for those who had ever been exposed and 2.2 (95% CI=1.0, 4.9) for those with 40 or more smoke-years of exposure after adjustment for age, race, and education. Similar odds ratios were observed for exposure to smoking by husbands and other household members in adult life (Table 7.5). Exposure to ETS from mothers, fathers, and siblings was associated with an increased risk of lung cancer, although none of the individual increases in risks were statistically significant. Stockwell et al. (1992) also considered ETS exposure from different sources during childhood/adolescence in terms of years of exposure. Women who experienced 22 years or more of ETS exposure from all household members combined during childhood/adolescence showed a significantly elevated OR for lung cancer (2.4, 95% CI=1.1,5.4) (Table 7.6). When ETS exposure from both childhood/adolescence and adulthood (i.e., from husbands and other household members) were considered jointly, women who reported 40 or more years of exposure experienced an elevated risk of lung cancer (OR=2.3, 95% CI=1.1, 4.6) compared to women who had fewer than 22 years of exposure (data not shown). These investigators noted that there was no statistically significant association in this study between ETS exposure at work or during social activities and risk of lung cancer (actual results were not presented in Stockwell et al., 1992).

The elevated risks associated with ETS exposure (during childhood/adolescence, adulthood, and all lifetime combined) were observed for all lung cancer cell types; the risk was stronger for cell types other than adenocarcinoma of the lung. Analysis by respondent type showed that the risk estimates for ETS exposure varied by the source of case information. For example, ETS exposure from husbands was a stronger risk factor for lung cancer when the respondents were the case patients (OR=3.1, 95% CI=0.9, 10.6) or their husbands (OR=3.1, 95% CI=0.7, 13.7). When the surrogate respondent was a family member other than the patient’s husband, ETS exposure was not associated with elevated risk (OR=0.9, 95% CI=0.4, 1.9).

It should be noted that the distribution of study subjects by ETS exposure was not presented; only the odds ratios were presented. The 'unexposed' reference category was comprised of individuals with no household ETS; presumably this same reference category was used in all analyses for cases and controls.

Brownson et al. (1992) conducted a population-based case-control study of women in Missouri (Table 7.4). Females aged 30 to 84 years who were diagnosed with primary lung cancer between January 1986 and June 1991, and were identified from the Missouri Cancer Registry, were considered eligible. Population controls were identified from a sample of the state driver's license files and Health Care Finance Administration listings. The case group included both lifetime nonsmokers and ex-smokers who had stopped smoking at least 15 years before diagnosis or who had smoked less than 1 pack-year. The definition of lifetime nonsmokers was not specified explicitly. The control group was matched by age group to case patients at about a 2 to 1 ratio. Tissue slides were reviewed to confirm the histologic classification for 468 (76%) of the 618 lung cancer cases.

The response rate was 95% for cases and 75% for controls, nonsmokers and ex-smokers combined. Information was collected on a total of 618 lung cancer cases of whom 432 were lifetime nonsmokers and 186 were ex-smokers. Of the lung cancer patients, 402 interviews were conducted with surrogate respondents and 216 interviews were with the lung cancer patients themselves. A total of 1400 control subjects were interviewed, all of whom were self-respondents; 1166 controls were lifetime nonsmokers.

All case and control interviews were conducted by telephone at which time the nonsmoking status was verified. Questions on ETS exposure pertained to exposures in both childhood (17 years and younger) and adult life (18 years and older). For each time period, respondents were questioned about the source of exposure (e.g., a parent or spouse) including both household and workplace exposure. After an individual source was determined, a series of detailed questions were asked on the type of tobacco used, duration of exposure, intensity of exposure, and average number of hours per day exposed. In the analyses restricted to lifetime nonsmokers, adjustment included age and history of previous lung diseases. Although initially examined, adjustment was not made for dietary beta-carotene and dietary fat because these factors did not confound these associations in this study.

In an analysis restricted to lifetime nonsmokers, there was no increase in risk associated with "ever-exposed" to spousal ETS (adjusted OR=1.0, 95% CI=0.8, 1.2) or exposure to fewer than 40 pack-years (see Table 7.5). However, analysis of the highest category of exposure to spouses' smoking (greater than 40 pack-years) yielded an OR of 1.3 (95% CI=1.0, 1.7) (Table 7.5). Analyses by histologic type showed the largest increase in risk for other/ mixed cell types and for small cell carcinomas, but these results were for lifetime nonsmokers and ex-smokers combined. Results were not presented separately for self-respondents and surrogate respondents. There was no association between risk of lung cancer and ETS exposure from parents (adjusted OR=0.7, 95% CI=0.5, 0.9) or other household members (adjusted OR=0.8, 95% CI=0.6, 1.1) during childhood (Table 7.6). These investigators also noted that there was no overall elevated lung cancer risk in this study associated with any ETS exposure in the workplace. However, lifetime nonsmokers showed an increase in risk at the highest quartile of workplace ETS exposure (OR=1.2, 95% CI=0.9, 1.7) (Table 7.7). Although the extent of exposure among the highest quartile of workplace was not specified, this OR is similar to the U.S. EPA report’s risk estimate for spousal smoking obtained from the meta-analysis.

Fontham et al. (1991, 1994) conducted a population-based case-control study of women in five geographic areas in the U.S. (New Orleans, Louisiana; Atlanta, Georgia; Houston, Texas; Los Angeles County, California; and the San Francisco Bay Area, California) (Table 7.4) (referred to as the U.S. multicenter study). Eligible cases included women with microscopically confirmed primary carcinoma of the lung that were diagnosed between December 1, 1986, and November 30, 1988, among residents of Atlanta and Houston, and during 2 additional years, 1989 and 1990, among residents of New Orleans, Los Angeles County, and San Francisco Bay Area. Additional eligibility criteria included age at diagnosis (20 to 79 years), language (English, Spanish, Chinese), history of previous cancer (none), and lifetime non-tobacco use (fewer than 100 cigarettes smoked and no use of any other form of tobacco for more than 5 months). One pathologist independently reviewed and confirmed histologic classification of 85% of the lung tumors in this study.

A population-based control group was selected by random digit dialing and supplemented by random sampling from the U.S. Health Care Financing Administration files for women 65 years and older. Controls were frequency matched to cases on race and age in a 2 to 1 ratio of controls to cases and met the same residence, language, and tobacco use criteria as cases. In-person interviews were completed for 665 of 800 incident lung cancer cases and 1278 of 1826 population controls; the respective response rate was 83% and 70%. The proportion of interviews conducted with self-respondents was 63% for lung cancer patients and 100% for controls. The considerably higher percentage of self-respondents in this study compared to the studies conducted by Stockwell et al. (1992) and Brownson et al. (1992) may be due to the more rapid identification of patients and thus contact of lung cancer cases in this multicenter study.

The lifetime nonsmoking status of study subjects was confirmed using a multistep procedure which included checking: 1) medical records, 2) with physicians’ offices, 3) at the time of contact to set up the interview, and 4) at the beginning of the interview. In addition, the subjects' current nonsmoking status was corroborated by measurement of urinary cotinine levels. Cotinine, a sensitive and specific biologic marker of recent tobacco exposure (Haley et al., 1983) was measured on 81% of self-respondent cases and 83% of controls. Levels of urinary cotinine/creatinine exceeding 100 ng/mg were found in 0.6% of cases and 2.3% of controls, indicating a low percentage of misclassification of smokers as nonsmokers (Fontham et al., 1994).

The in-person interviews followed an extensive structured questionnaire designed to obtain information on household, occupational, and other exposures to ETS during each subject's lifetime, as well as other exposures associated with lung cancer. Exposure to ETS was examined by source during childhood (father, mother, and other household members who lived in the home for at least 6 months) and during adult life (spouse, other household members, occupational, and social exposures).

Spousal smoking was associated with a statistically significant increased risk of lung cancer; adjusted ORs of 1.29 (95% CI=1.04, 1.60) for ever exposed to spouses' smoking and 1.79 (95% CI=0.99, 3.25) (p for trend=0.03) for 80 or more pack-years to spouses' smoking were observed (Table 7.5). Exposure to other sources of ETS during adult life were also associated with an increased risk of lung cancer. Adjusted ORs of 1.39 (95% CI=1.11, 1.74) for ever exposed to ETS at the workplace and 1.86 (95% CI=1.24, 2.78) for 31 or more years of exposure at the workplace were observed (Table 7.7). In addition, increased risks were associated with ETS exposure in social settings (see section 7.2.4.3). When all sources of ETS exposure during adult life were considered jointly as years of exposure, women with 48 years or more of exposure showed an OR of 1.74 (95% CI=1.14, 2.65) compared with women with no ETS exposure (data not shown). The increased risks associated with ETS exposure from spouses, at the workplace, and other social settings were observed for adenocarcinomas as well as other histologic types of lung cancer.

The findings for ETS exposure were similar when the analysis was restricted to self-respondents only. For example, among self-respondents only, an OR of 1.67 (95% CI=1.03, 2.70) was found for women with 48 years or more of exposure for all sources combined in adult life compared with women with no exposure (the OR was 1.74 for all respondents combined) (data not shown). These results for ETS exposure were observed after adjustment for age, race, study area, education, intake of fruits and vegetables and use of supplemental vitamins, dietary cholesterol, family history of lung cancer, and employment in high-risk occupations.

In this study, ETS exposure during childhood/adolescence from father, mother, or other household members was not associated with risk of lung cancer. The OR for any childhood exposure to ETS (i.e., any household member) was 0.89 (95% CI=0.72, 1.10) (Table 7.6) (data from table 4 of Fontham et al. (1994)). However, there was some suggestion that the risk associated with adult ETS exposure varied according to childhood ETS exposure. Significantly elevated risks associated with adult ETS exposures were observed in women with and without childhood exposures. The elevations in risk for women exposed during childhood were twice as high as those without childhood exposures. For example, at the highest level of ETS exposure (48 adult smoke-years or more), the authors reported an adjusted OR of 3.25 (95% CI=1.42, 7.46) among women reporting childhood exposures compared to 1.77 (95% CI=0.98, 3.19) for those reporting no childhood exposure (data not shown).

Kabat et al. (1995) conducted a hospital-based case-control study of women and men between 1983 and 1990 as part of a long-standing study of tobacco-related cancers. This study was carried out in six hospitals located in four U.S. cities (New York City, New York; Chicago, Illinois; Detriot, Michigan; and Philadelphia, Pennsylvania). Newly diagnosed, histologically confirmed cases of primary cancer of the lung were ascertained in the collaborating hospitals. For each case who was enrolled, up to three control patients who were lifetime nonsmokers matched on age (+ 5 years), sex, race, hospital, and date of interview (within 2 months) were interviewed. Control patients were admitted for various cancer and noncancer outcomes. About 30% of the controls were diagnosed of cancer of the stomach/intestine, genitourinary tract, or lymphatic and hematopoietic system, cancer sites which may be positively associated with tobacco use (see sections 7.3.3, 7.4.2, and 7.4.4). Thus, the ETS exposure among some controls may be higher than the general population, leading to a bias towards the null.

Subjects were considered lifetime nonsmokers if they had never consumed as much as 1 cigarette per day for a year, or had smoked fewer than 365 cigarettes over their lifetime. In the structured interview, detailed questions regarding the initiation of smoking early in life were included and provided a basis for excluding ex-smokers who quit decades prior to diagnosis but had smoked more than this minimum amount. The proportion of never-smokers among all lung cancer cases in this study was 3 percent in males and 8 percent in females.

All subjects were interviewed in person in the hospital. The questionnaire included a detailed history of exposure to ETS, during childhood and adult life. Questions were also asked about adult ETS exposures inside and outside the home (at work, in cars and other forms of transportation, and in social settings). Interviews were conducted with 41 male and 69 female never-smoking lung cancer cases and 117 male and 187 female never-smoking controls.

There were no significant associations between spouses’ smoking and risk of lung cancer in male (OR=1.60, 95% CI=0.67, 3.82) or female (OR=1.08, 95% CI=0.60, 1.94) subjects (Table 7.5). We calculated the OR for lung cancer in males and females combined to be 1.19 (95% CI=0.76, 1.87) in association with spousal ETS exposure. Wives’ smoking 11+ cigarettes/day was associated with a significant increased risk (OR=7.48, 95% CI=1.35, 41.36) of lung cancer in men (Table 7.5). However, this result was based on small numbers and thus unstable, and a similar result was not observed in women associated with their husbands’ smoking. For males and females combined, we calculated the OR for having a spouse who smoked 11+ cigarettes/day to be 1.57 (95% CI=0.81, 3.07). The OR for lung cancer associated with spouses who smoked in the bedroom was slightly higher than any smoking by spouses, but this association was not statistically significant in males (OR=5.02, 95% CI=0.72, 35.01) or females (OR=1.09, 95% CI=0.49, 2.42) (data not shown)(we calculate the crude OR for males and females combined to be 1.20, 95% CI=0.6, 2.4).

Results for any household ETS exposure during adult life were similar to the results described above for spousal ETS exposure; household exposure was not significantly associated with risk of lung cancer (Table 7.6). The exception was that, among males, there was a statistically significant increased risk (OR=4.15, 95% CI=1.34, 12.87) of lung cancer associated with 2 or more smokers in the adult household, but this was not observed among females (OR=0.94, 95% CI=0.34-2.63).

Sources of ETS exposure outside of the home during adult life were also evaluated, including ETS exposure at the workplace, in social situations, and inside cars. Workplace ETS exposure was not associated with increased risk of lung cancer in males or females (Table 7.7) in this study. There were small increased risks for lung cancer associated with ETS exposures in social situations and inside cars (see 7.2.4.3). The elevated risk associated with ETS exposure inside cars was statistically significant in an analysis which combined male and female subjects (see 7.2.4.3).

Exposure to ETS during childhood was not associated with any increased risk in males (OR=0.90, 95% CI=0.43,1.89), but it was associated with an increased risk in females of borderline statistical significance (OR=1.55, 95% CI=0.95,2.79) (Table 7.6). There were no significant dose-response relationships between number of smokers in childhood households and risk of lung cancer in male or female subjects in this study.

The results from these four recent U.S. studies are compatible with the pooled estimate of the U.S. EPA (1992) report, which found a summary OR of 1.19 (90% CI=1.04, 1.35) for ever exposed to ETS from spouses (for U.S. studies). Results from the largest population-based study, the U.S. multicenter study (OR=1.29, 95% CI=1.04, 1.60, for ever exposed) (Fontham et al., 1994) were closest to the pooled estimate from the U.S. EPA report. Of the two other population-based studies, the association found in the Florida study (Stockwell et al., 1992) was stronger (OR=1.6, 95% CI=0.8, 3.0; although it did not achieve statistical significance except for the highest exposure category: OR=2.4, 95% CI=1.1, 5.3), and that from the Missouri study (Brownson et al., 1992) was weaker (overall OR=1.0, 95% CI=0.8, 1.2; for highest exposure category of spousal smoking, OR=1.3, 95% CI=1.0, 1.7) than the pooled estimate result. Although the authors of the fourth study, the hospital-based case-control study (Kabat et al., 1995), interpreted their findings (analyzing men and women separately) to be unsupportive of an association between ETS exposure and risk of lung cancer, the odds ratio we calculated from their results (OR [for males and females combined] = 1.19, 95% CI=0.76, 1.87), though not statistically significant, was in fact very similar to the pooled estimate from the U.S. EPA report. In addition, positive increasing trends in risk of lung cancer in nonsmokers were observed for increasing ETS exposure indices in all three of the population-based studies (Table 7.5). The concordance in these study results gives further credibility to the finding of a causal association between spousal ETS exposure and risk of lung cancer described in the U.S. EPA (1992) report.

The sample sizes of the three population-based U.S. studies (Stockwell et al., 1992; Brownson et al., 1992; Fontham et al., 1994) were considerably larger than previously published case-control studies in the U.S. (Correa et al., 1983; Buffler et al., 1984; Kabat and Wynder 1984; Dalager et al., 1986, Wu et al., 1985, Garfinkel et al., 1985, Humble et al., 1987; Brownson et al., 1987; Janerich et al., 1990). Spousal ETS exposure was not associated with a significant increased risk of lung cancer in males and females in the Kabat et al. (1995) study. However, this study was considerably smaller than the other three U.S. studies published in the 1990s and had limited statistical power to detect a significant association.

Another important feature of the four post-1991 studies is that they addressed many of the criticisms (Mantel, 1983; Lee, 1986 and 1989; Katzensten, 1992) directed at previous studies of ETS exposure and lung cancer. Although the extent to which these criticisms were addressed in each of the four studies varied, these concerns were addressed collectively in these studies. Specifically, concerns regarding selection bias, misclassification bias of smokers as lifetime nonsmokers, misclassification of some non-lung cancers as lung cancers, misclassification of ETS exposure, and the lack of adjustment of potential confounders were addressed.

The three population-based studies were careful to minimize the possibility of selection and misclassification biases. Selection bias associated with cases from selected hospitals is eliminated since, in all three population-based studies, lung cancer patients were identified from the cancer registries and hospitals covering a specific study area. The use of population-based controls instead of other patients as controls is also advantageous, since ETS exposure of patients with certain diagnoses may be higher and not representative of the exposure distribution of the source population from which cases were drawn. In addition, the U.S. multicenter study (Fontham et al., 1991) examined the issue of differential recall between lung cancer cases and controls by interviewing colon cancer patients as a second control group during the first three years of the study. The findings on ETS exposure were comparable when lung cancer patients were compared to population controls and to colon cancer controls, suggesting that recall bias resulting from having a diagnosis of cancer could not explain the observed association with ETS.

Another source of misclassification bias that is of concern (Wald et al., 1986; Lee, 1989) pertains to the misclassification of smokers as nonsmokers. In two of the four studies (Fontham et al., 1994; Stockwell et al., 1992), the definition of lifetime nonsmokers was limited to individuals who had smoked fewer than 100 cigarettes and had no more than 6 months of tobacco use in their lifetime. In one study (Kabat et al., 1995), subjects were considered lifetime nonsmokers if they had never consumed as much as 1 cigarette per day for a year, or had smoked fewer than 365 cigarettes over their lifetime. Both the U.S. multicenter study (Fontham et al., 1994) and the Florida study (Stockwell et al., 1992) used multiple sources of information to verify the lifetime nonsmokers' status. In addition, the U.S. multicenter study corroborated the subjects' self-reported current nonsmoking status using the urinary cotinine level. These results showed a very low percentage of cases (0.6%) and controls (2.3%) had levels of urinary cotinine exceeding 100 ng/mg, suggesting minimal misclassification of smokers as nonsmokers (Fontham et al., 1994). Although the urinary cotinine/creatinine concentration only assesses current smoking (there are currently no biomarkers that allow assessment of past tobacco exposure), these results provided an additional verification of the current nonsmoking status.

Misclassification of lung cancer is also minimized by the requirement of microscopic diagnosis (Stockwell et al., 1992; Fontham et al., 1994; Kabat et al., 1995) and an independent review of diagnostic material (Brownson et al., 1992; Fontham et al., 1994). In the three population-based studies with data by cell type (Stockwell et al., 1992; Brownson et al., 1992; Fontham et al., 1994), adenocarcinoma of the lung was the predominant cell type of lung cancer in nonsmoking women, accounting for over 60% of the lung tumors.

Because of the high fatality rate of lung cancer, all 3 U.S. population-based studies interviewed surrogate respondents to obtain information on a percentage of lung cancer cases who could not participate because they were too ill or were deceased. In all three studies, controls were self-respondents. The percentage of lung cancer self-respondents was considerably higher for the U.S. multicenter study (63%) compared with the other two U.S. studies (33% for the Florida and 34% for the Missouri study). Since a surrogate's knowledge of the ETS exposure of an index subject is variable, dependent on their relationship and the exposure period of interest, it is likely that the quality of information on ETS exposure is higher in studies in which a high percentage of interview is conducted with self-respondents. On the other hand, the use of surrogate respondents was avoided in the U.S. hospital-based study since all interviews were conducted with the lung cancer cases and hospital patient controls while the subjects were still in the hospital (Kabat et al., 1995).

Another criticism of previous studies of ETS exposure and lung cancer is that a small increased risk associated with ETS exposure may be due to lack of adjustment for potential confounding factors. In particular, it has been suggested that nonsmokers living with smokers have lower dietary intakes of specific micronutrients (Koo et al., 1988; Hebert and Kabat, 1990; Sidney et al., 1989; Le Marchand et al., 1991; Matanoski et al., 1995), including beta-carotene, which may be protective for lung cancer. However, there is little evidence of confounding by dietary factors in the U.S. multicenter study (Fontham et al., 1994) or in a study conducted in Greece (Kalandidi et al., 1990). In fact, similar trends of increased risk of lung cancer associated with increasing duration of exposure were observed at all levels of dietary factors (including intake of fruits and vegetables, supplemental vitamin use, and dietary cholesterol) (Fontham et al., 1994). Other factors including employment in high-risk occupations (Fontham et al., 1994) and previous lung diseases (Brownson et al., 1992; Wu et al., 1995) were examined and they did not confound the association of ETS exposure and lung cancer.

Because of the importance of obtaining a comprehensive measure of lifetime ETS exposure (Cummings et al., 1989), all four U.S. studies included questions to assess ETS exposure at home (from spouses, parents, and other household members during childhood and adult life), at the workplace and in other social settings. However, the exact questions asked and the level of detail obtained varied in these studies. Only a subset of the studies published prior to 1991 included questions on ETS exposures other than from spouses.

7.2.4.1 ETS Exposure From Parents and Other Household Members (Other Than Spouses)

Table 7.6 summarizes the case-control studies conducted since 1981 in the U.S. (n=7) and outside of the U.S. (n=7) which included questions on ETS exposure from household members other than spouses, represented mainly by exposure from parents during childhood, but also including other household members during childhood and adult life. The study by Akiba et al. (1986) which reported 'no association' was not included in Table 7.6 since no information on the association or the distribution of subjects by exposure status was provided.

Among the U.S. studies, the strongest evidence for an effect of parental smoking is from studies conducted by Janerich et al. (1990) and Stockwell et al. (1992). In the study by Janerich et al. (1990), exposure during childhood up to age 21 accounted for about one-third of the lifetime duration (expressed in smoker-years) of ETS exposure. The highest level of childhood exposure (25 or more smoker-years) was associated with a statistically significant increased risk (OR=2.07, 95% CI=1.16, 3.68), although there was no statistically significant elevated risk with 1-24 years of exposure. In the study by Stockwell et al. (1992), exposure to ETS from mothers, fathers, and siblings during childhood/adolescence was associated with a 10 to 70% increase in risk. Women who experienced 22 years or more of exposure to ETS from all household members combined during childhood/adolescence showed a significantly elevated risk of 2.4 (95% CI=1.1, 5.4) (Table 7.6). On the other hand, risk of lung cancer in nonsmokers was not associated with ETS exposure during childhood in the U.S. multicenter study (Fontham et al., 1994), the Missouri study (Brownson et al., 1992), the U.S. hospital-based study (Kabat et al., 1995) and a small study conducted in Los Angeles County (Wu et al., 1985). However, in the U.S. multicenter study, subjects who were exposed to ETS exposure during both childhood and adult life showed the highest increase in risk of lung cancer (Fontham et al., 1994). In a hospital-based study conducted in the 1970s (Kabat and Wynder, 1984) and a subsequent one conducted in the 1980s (Kabat et al., 1995), smoking by family members during adult life was not associated with risk of lung cancer in nonsmoking males and females.

In two studies conducted in Japan (Shimizu et al., 1988; Sobue, 1990), an increased risk of lung cancer was associated with mothers’ smoking; the result was statistically significant in one study (Shimuzu et al., 1988) but not the other (Sobue, 1990). A significantly increased risk of lung cancer was also associated with smoking by the father-in-law in one Japanese study (Shimizu et al., 1988). In Shanghai, China (Gao et al., 1987) and in Northern China (Wu-Williams et al., 1990), exposure to ETS during childhood did not differ significantly between lung cancer cases and controls. In a study conducted in Hong Kong, risk of lung cancer in nonsmokers was increased in households with smokers, although there was not a smooth trend of increasing risks with increasing number of smokers in the household (Koo et al., 1987). In Sweden, no association between parents’ smoking and risk of lung cancer was reported in one study (Pershagen et al., 1987), whereas in another study, a statistically nonsignificant 3-fold increased risk of lung cancer was found for mothers' smoking (Svensson et al., 1989).

Quality of information on parents' smoking (or other household members) during childhood may be compromised in some studies, particularly those in which this information is provided by surrogate respondents. Although there is generally good agreement of responses on ETS exposure when subjects themselves were asked on two different occasions whether specific household members smoked, the level of agreement diminished on quantitative aspects of smoking by household members (Pron et al., 1988; Coultas et al., 1989; Brownson et al., 1993a). Studies which show high concordance on the reporting of exposure to ETS during childhood and parents' smoking habits (Coultas et al., 1989; Brownson et al., 1993a) were based on responses obtained from the subjects themselves. The degree of agreement when the responses on smoking habits of the other household members are provided by surrogate respondents is not known. The fact that exposures from household members other than spouses are reported less reliably may partially explain the inconsistent results regarding the association between the risk of lung cancer and ETS exposure from these household members (i.e., other than spouses), and the failure in most studies to observe stronger associations with exposure from household members other than from spouses.

7.2.4.2 Workplace ETS Exposure

Table 7.7 summarizes case-control studies which included questions on ETS exposure at the workplace. Indicators of workplace ETS exposure varied (the actual questions asked were not provided). In some studies, the indicators of workplace ETS exposure were limited to the most recent job or the last job (Kabat and Wynder, 1984; Shimizu, 1988; Kalandidi et al., 1990; Brownson et al., 1992), at other specific times (Garfinkel et al., 1985), or the timing of the question was not specified (Lee et al., 1986; Stockwell et al., 1992). In one study, number of smokers at work (lifetime) and amount of time working with smokers was assessed (Janerich et al., 1990). In other studies, questions were asked regarding ETS exposure at each workplace of at least three months (Koo et al., 1987) or the last four jobs of at least one year duration (Kabat et al., 1995). In three other studies, lifetime occupational history was obtained and exposure to ETS was assessed for each job (Wu et al., 1985; Wu-Williams et al., 1990; Fontham et al., 1994).

Studies in which the assessment of workplace exposure to ETS was complete (covering all jobs) with considerable ETS exposure of subjects in the studies are generally supportive of an association between workplace ETS exposure and risk of lung cancer (Wu et al., 1985; Wu-Williams et al., 1990; Fontham et al., 1994). In particular, results from the U.S. multicenter study (Fontham et al., 1994) suggested a trend of increasing risks with increasing duration of ETS exposure at the workplace. Compared to women who had no ETS exposure at the workplace, women who reported exposure for 1-15, 16-30, and 30 or more years showed adjusted odds ratios of 1.30, 1.40, and 1.86, respectively (p for trend=0.001) (Table 7.7). In a subsequent analysis which selected workers only and adjusted for other adult ETS exposure sources, the RRs associated with workplace exposure were modestly enhanced (Reynolds et al., 1996). The overall odds ratios associated with any reported workplace exposure increased from 1.39 in the earlier analysis to 1.56 (95% CI=1.21-2.02) and the corresponding point estimates for 1-15, 16-30, and 30 or more years of exposure were likewise elevated; the adjusted odds ratios were 1.46, 1.58, and 2.08, respectively. Occupational exposure to carcinogens is an important confounder for lung cancer in nonsmokers, and the U.S. multicenter study (Fontham et al., 1994) is the only one which adjusted for such exposures.

In addition to the incomplete assessment of exposure to ETS at the workplace in some studies, respondents, particularly surrogate respondents, may be less able to provide information on the subjects' exposure to ETS at the workplace. In a study in which a test-retest design was used to examine the reliability of passive smoke histories reported in personal interviews, self-respondents more reliably reported residential exposure than exposure at work (Pron et al., 1988). This may be a particularly important problem in studies in which the proportion of surrogate respondents was high (Brownson et al., 1992; Stockwell et al., 1992).

Despite some of the above mentioned difficulties in obtaining histories of lifetime ETS exposure at the workplace, there is reason to believe this source of ETS exposure also increases the risk of lung cancer, as does ETS exposure from spouses. The workplace has been a major source of ETS exposure outside the home (Cummings et al., 1989 and 1990; Emmons et al., 1992; Siegel, 1993), although the relative importance of workplace ETS exposure may be declining in California as the result of increasing restrictions on smoking in the workplace. In the International Agency for Research on Cancer (IARC) 10-country collaborative study which correlated urinary cotinine levels to self-reported recent exposure to ETS at home (from spouses), in the workplace, and other social settings, Riboli et al. (1990) found that exposure to ETS at the workplace was a significant predictor of cotinine levels, similar to ETS exposure from spouses.

7.2.4.3 ETS Exposure in Other Settings

Two of the four U.S. studies published since 1991 (Fontham et al., 1994; Kabat et al., 1995) also asked questions about ETS exposure in social settings (other than the workplace) or in modes of transportation. In the U.S. multicenter study, increased risks were associated with ETS exposure in social settings. Women who were exposed for 1-15, 16-30, and >30 years at other social settings compared to no exposure showed adjusted ORs of 1.45, 1.59, and 1.54, respectively (p for trend=0.002) (Table 6 of Fontham et al., 1994). In the U.S. hospital-based study, associations with ETS exposure in social situations and risk of lung cancer were not statistically significant in males (OR=1.39, 95% CI=0.67, 2.86) or females (OR=1.22, 95% CI=0.69, 2.15) (Table 2 of Kabat et al. (1995); we calculated the OR for males and females combined to be 1.26 (95% CI=0.81, 1.95). ETS exposure in cars was associated with nonsignificant increased risks of lung cancer in both males (OR=1.55, 95% CI=0.63, 3.78) and females (OR=1.84, 95% CI=0.96, 3.53) in the Kabat et al. (1995) study. Although the risks for males and females considered separately were not significantly different from controls, the risk we calculated in males and females combined was significantly elevated (OR=1.73 (95% CI=1.03, 2.92). No male cases were exposed to ETS in other modes of transportation, whereas there was a significant excess of female cases compared to female controls who reported such exposures (OR=5.17, 95% CI=1.46, 18.24). We calculated that ETS exposure in other modes of transportation was associated with an OR of 2.23 (95% CI=0.83, 5.99) for lung cancer in males and females combined, in the Kabat et al. (1995) study.

Despite the compelling biologic plausibility of an effect of ETS exposure on risk of lung cancer, detection of an effect has been problematic because a small excess in risk is difficult to establish in a single epidemiologic study. The U.S. EPA (1992), NRC (1986) and Surgeon General (U.S. DHHS, 1986) all undertook comprehensive reviews of the literature and determined on the basis of the overall evidence that ETS exposure causes lung cancer. Since the publication of the most recent authoritative review of lung cancer and ETS exposure (U.S. EPA,1992), three large U.S. population-based studies (Stockwell et al., 1992; Brownson et al., 1992; Fontham et al., 1991 and 1994) and a smaller hospital-based case-control study (Kabat et al., 1995) have been published. The three population-based studies were designed to and have successfully addressed many of the weaknesses (i.e., small sample size, possible selection bias, possible misclassification biases, inadequate adjustment for potential confounders) for which the previous studies on ETS and lung cancer have been criticized. Results from these studies are consistent with the conclusions of the U.S. EPA (1992), NRC (1986) and Surgeon General (U.S. DHHS, 1986) reports. Each of the three population-based studies show a statistically significant increased risk of lung cancer in nonsmokers associated with long term exposure to ETS as well as increasing risk with increasing ETS exposure. The smaller hospital-based study lacked the statistical power to find the effect observed in the other studies. However, though not statistically significant, the OR we calculated for male and female combined in this study was similar to that reported by the U.S. EPA. Taken together, the recent studies provide additional evidence that ETS exposure is causally associated with lung cancer. The consistency of the findings in the four recent studies and the meta-analysis result of the U.S. EPA indicates about a 20% increased risk of lung cancer in nonsmokers.

Active smoking is firmly established as a causal factor for cancers of the lung, larynx, oral cavity, esophagus, bladder, and nasal sinus cavity; in addition, evidence exists which suggests that smokers are at increased risk of kidney and cervical cancer. As reviewed above, the role of ETS exposure and risk of cancers in nonsmokers has been investigated mainly for lung cancer (U.S. DHHS, 1986; NRC, 1986; U.S. EPA, 1992). There are some data on the role of ETS for other cancer sites, including cancers of nasal sinus cavity, cervix, and bladder (U.S. DHHS, 1982 and 1989; IARC, 1986).

7.3.1.1 Active smoking and nasal sinus cancer

Cancers of the nasal cavity and paranasal sinuses are extremely rare, accounting for 0.2 per cent of all invasive incident cancers and 1.4 percent of all newly diagnosed respiratory cancers in the U.S. Use of tobacco products, various occupational exposures (e.g., wood dust), and history of nasal polyps, have been implicated as risk factors for these tumors (Elwood et al., 1981; Brinton et al., 1984; Hayes et al., 1987; Strader et al., 1988; Zheng et al., 1992). Although the risk associated with any use of tobacco is modest (OR about 1.5), up to a 5-fold increased risk has been observed with heavy smoking (Elwood et al., 1981). The evidence suggests that the effect of smoking, particularly current or recent tobacco use, is stronger for squamous cell carcinoma than for other cell types (mainly adenocarcinomas) of nasal sinus cancer (Elwood et al., 1981; Brinton et al., 1984; Hayes et al., 1987; Strader et al., 1988; Zheng et al., 1992). The proportion of squamous cell nasal sinus cancers included in the different studies may influence the overall strength of the relationship between active smoking and all nasal sinus cancers combined. Studies which did not find a significant association between active smoking and nasal sinus cancer were generally small studies (i.e., <50 cases and controls) (Tola et al., 1980; Merler et al., 1986), or had included few squamous cell carcinomas of the nasal sinus. For example, the study by Merler et al. (1986) included less than 20% squamous cell carcinomas compared to at least 40% of this cell type in other studies finding a positive association with smoking (Elwood et al., 1981; Brinton et al., 1984; Hayes et al., 1987; Strader et al., 1988; Tola et al., 1980; Zheng et al., 1992).

7.3.1.2 ETS and nasal sinus cancer

The role of ETS exposure in the etiology of nasal sinus cancer in nonsmokers has been investigated in one cohort and two case-control studies (Table 7.8).

Using data from a Japanese prospective study (see Section 7.1 for detailed description), Hirayama (1983) reported an increased risk of para-nasal sinus cancer (based on 28 nasal sinus cancer deaths) among nonsmoking women exposed to husbands' smoking. Relative risks increased with amount husbands smoked: compared to women married to nonsmokers, the RR was 1.7 (95% CI=0.7,4.2), 2.0 (95% CI=0.6,6.3), and 2.6 (95% CI=1.0, 6.3), for women whose husbands smoked 1-14, 15-19, and 20+ cigarettes per day respectively, when husbands' age and occupation were adjusted for. However, active smoking was not associated with nasal sinus cancer in this study; the OR associated with active smoking was 0.9 (90% CI=0.5-1.4) for males and females combined (Hirayama, 1990). Cell type distribution of nasal sinus cancer in nonsmokers and smokers was not available in this Japanese cohort study.

The second study was conducted by Fukuda and Shibata (1988; 1990) in Japan, using a case-control study design. The 1988 report presented preliminary findings and the 1990 report included results on 169 (125 men and 44 women) squamous cell maxillary sinus cancer cases and 338 controls (250 men and 88 women). Controls were selected from the general population. All subjects were interviewed directly. Nine of 125 male cases and 48 of 250 male controls had never smoked. Active smoking was a significant risk factor in men; the RR was 4.6 for smoking >39 cigarettes per day compared to nonsmokers. Based on a small number of nonsmoking men, exposure to ETS was associated with a small, nonsignificant increased risk of nasal cancer. Most of the female cases and controls in this study were nonsmokers (35 of 44 cases and 74 of 88 controls had never smoked). Active smoking was associated with a nonsignificant increased risk of nasal cancer in women. Among nonsmoking women, domestic exposure to ETS, represented by the number of smokers in the household, was a significant risk factor. Compared to nonsmoking women with no reported ETS exposure, nonsmoking women who reported 1, and 2 or more smokers in the household showed RRs of 1.4 and 5.7, respectively (p for trend = 0.02). We calculated the 95% confidence intervals for the point estimates 1.4 and 5.7 as 0.6-1.5 and 1.7-19.4, respectively. The OR associated with any passive smoke exposure (i.e., none versus any exposure) is 1.96 (95% CI=0.8-4.5). Information on duration or intensity of ETS exposure was not reported. The effect associated with passive smoking persisted with adjustment for other risk factors including sinusitis and/or polyps, nasal trauma, and woodworking.

The third study was a case-control analysis of cancer of the nasal cavity and sinuses among white men in the US using data from the 1986 National Mortality Followback Survey (Zheng et al., 1993). The study included a total of 147 cases (76 maxillary sinus, 11 nasal cavity, 4 auditory and middle ear, 56 other accessory sinuses cancer) and 449 controls who died of other causes. All information was obtained from a surrogate who responded to a mailed questionnaire. There was an increased risk of nasal cancer among cigarette smokers, with a nearly two-fold increased risk among heavy or long-term smokers for all nasal cancer sites. Compared to nonsmokers, heavy smokers showed an OR of 2.7 (95% CI=1.2-6.4) for maxillary sinus cancers and an OR of 1.3 for other nasal cancer (95% CI=0.5-3.3). Twenty-eight cases and 99 controls had never smoked. Among nonsmokers, more cases than controls had a wife who smoked cigarettes (OR = 3.0, 95% CI = 1.0 - 8.9), but the authors stated there was not a smooth trend of increasing risks as the number of cigarettes smoked by the spouse increased (data on dose-response were not presented). The 3-fold risk associated with having a wife who smoked is somewhat surprising since more than half (15 of 28) of the tumors in nonsmokers were other nasal sinus cancer and this subgroup was less strongly associated with active smoking. However, the histologic cell type of nasal sinus cancer among smokers and nonsmokers was not available in this study, making it difficult to make direct comparisons of findings in smokers and nonsmokers.

7.3.1.3 Summary

Existing studies consistently show a significant positive association between exposure to ETS and nasal sinus cancer in nonsmokers. The results have been observed in studies conducted in eastern and western countries, in males and females, in cohort and case-control study designs, and with some adjustment for possible confounders. The risks associated with ETS exposure ranged from 1.7 to 3.0.

Future studies need to confirm the magnitude of risk associated with ETS exposure, to characterize the risk by the source of ETS exposure (i.e., spouse, other household members, at work) and by timing of ETS exposure (current versus past exposure), and to establish the dose-response relationship. It is also important that future studies examine the association between ETS and nasal sinus cancer by histologic type and subsite of nasal sinus cancers, and the role of other potential confounders in the association. Studies designed to investigate the mechanism(s) of action of active smoking and ETS exposure will help to elucidate their respective roles in the development of nasal sinus cancer.

7.3.2.1 Active smoking and cervical cancer

Numerous epidemiologic studies conducted in different populations, of different age groups, and among different degrees of cervical lesions have provided supportive evidence that women who smoke cigarettes are more likely to develop cervical cancer than women who do not (Winkelstein 1990). The statistical association between active smoking and cervical cancer is reduced with adjustment for sexual activity variables (e.g., number of partners, age at first intercourse) or infection with human papillomarivus (HPV), which has been accepted as the sexually transmitted etiological factor in cervical cancer (Brinton, 1990; Schiffman et al., 1993; Munoz et al., 1994; zur Hausen, 1986). However, an association between smoking and cervical cancer/intraepithelial neoplasia (CIN) has been found in case-control studies which have been able to control for these behavioral risk factors (Buckley et al., 1981; Hellberg et al., 1983; Brinton et al., 1986; Clarke et al., 1982; La Vecchia et al., 1986; Becker et al., 1994).

In most studies, the excess risk of cervical cancer for smokers is about two-fold, with the highest risks generally observed for heavy or current smokers, suggesting that tobacco smoke may have a late-stage effect on cervical cancer development. The data also suggest that tobacco smoke may be a cofactor in the development of particularly high-grade CIN (Brinton and Hoover 1992; Schiffman et al., 1993) by acting with or enhancing other infectious agents, such as cervical HPV (zur Hausen, 1986; Burger et al., 1993) in the promotion of cervical neoplasia. A possible mode of action of tobacco smoke is to compromise immune function (Barton et al., 1988).

In addition to the epidemiological evidence, an association between smoking and cervical cancer is biologically plausible since carcinogens in tobacco smoke can be absorbed in the lung and transported to distant sites by the blood. Tobacco constituents, including cotinine and nicotine, have been detected in the cervical mucus of smokers (see below). Higher levels of DNA adducts in cervical biopsies of smokers compared to nonsmokers have also been reported (Simons et al., 1994) (see below). Among women with cervical dysplasia, higher levels of mutagenicity in the cervical mucus of smokers compared to nonsmokers have been found (Holly et al., 1986), although this result has not been observed in studies of women without cervical dysplasia (Schiffman et al., 1987; Holly et al., 1993).

7.3.2.2 ETS exposure and cervical cancer

The relationship between ETS exposure and cervical cancer was investigated in one cohort and three case-control studies (Table 7.9).

As part of the Japanese cohort study, Hirayama (1981) presented results on risk of cervical cancer in women by husbands' smoking habits. Based on 250 cervical cancer deaths in nonsmokers, the RRs were 1.15, and 1.14 for women whose husbands were ex-smokers or smoked 1-19 cigarettes/day, and >20 cigarettes/day, respectively, compared to women whose husbands were nonsmokers (p value for trend= 0.25). In the same study, women who ever smoked showed a high risk fo cervical cancer compared to nonsmokers (RR=1.6, 90% CI=1.3-1.9), and there was some suggestion of increasing risks with increasing amounts smoked (the RRs associated with smoking 1-9, 10-19, and 20+ cigarettes/day were 1.7 (90% CI=1.3-2.2), 1.3 (90% CI=1.0, 1.8) and 2.4 (90% CI=1.4-3.9), respectively (Hirayama, 1990)). The findings on active smoking and passive smoking were not adjusted for potential confounders including subjects' or husbands' sexual activity.

A case-control which provided some data on the role of ETS exposure (see Section 7.1.1 for details) and risk of cervical cancer in nonsmokers was a study on childhood and adult life ETS exposure and risk of various cancer outcomes (Sandler, 1985a). Because this study included different cancer outcomes, information typically obtained in studies of a specific cancer site (e.g., sexual activity in studies of cervical cancer) was not collected. There were a total of 518 cancer patients; 101 had cervical cancers of which 56 occurred in women who had never smoked. The 56 nonsmokers with cervical cancer were compared to 235 nonsmoking control women. Spouses' smoking habits were associated with an increased risk of cervical cancer in nonsmokers (OR =2.1, 95% CI =1.2, 3.9) after adjustment for age, race, education, and smoking habits of parents. In the same study, husbands' smoking also increased risk of cervical cancer in women who were smokers (OR=2.0, 95% CI = 0.9, 4.1); the effect was observed after adjustment for the above mentioned variables as well as personal smoking habits of women. Sandler et al. (1985b) also examined the association between parental smoking during childhood and risk of cervical cancer. Maternal smoking was not associated with risk of cervical cancer (OR = 0.66, 95% CI=0.19-2.29) whereas paternal smoking was associated with a statistically nonsignificant increased risk (OR = 1.67, 95% CI=0.81-3.45). The difference in results for mothers' versus fathers' smoking is likely due to chance; among controls the prevalence of mothers who smoked was low (11%) compared to fathers who smoked (43%).

A second case-control study on this subject was designed to investigate the role of active smoking and passive smoking in the etiology of cervical cancer. This study included 266 women with cervical cancer and 408 population controls, selected by random-digit dialing in Utah (Slattery et al., 1989). Eighty-one cases and 305 controls had never smoked. Women were asked whether they were exposed to "a lot, some, a little, or no" tobacco smoke inside or outside their homes and the number of hours of exposure per day, during the 5 years before interview. Among nonsmokers, ETS exposure inside and outside of the home was associated with a significantly increased risk with adjustment for potential confounders which included age, education, church attendance, and number of sexual partners of the woman. A 3-fold increased risk (OR=3.4, 95% CI=1.2-9.5) was observed for 3 or more hours of exposure per day. The increased risks associated with ETS exposure among nonsmoking women were comparable to the risks associated with active smoking in this study (Table 7.5). Although specific information on HPV infection or partners' sexual activity was not available, the effect of active smoking was strongest among women who had a few (none to 1) sexual partners (OR=14.2, 95% CI=9.2,38.9) for smoking) and weakest for women with 4 or more partners (this was the highest category of partners in this study) (OR=2.3, 95% CI=1.4, 3.9) for smoking). The authors interpreted this finding to suggest that cigarette smoking and presumably ETS exposure as risk factors for cervical cancer may be more important among women who have not experienced other major risk factors for this cancer (i.e., positive for HPV infection).

Another case-control study which was designed to examine the role of active and passive smoking included 103 CIN cases (40% CIN II, 60% CIN III) and 268 controls; 37 CIN cases and 170 controls had never smoked (Coker et al., 1992) (Table 7.9). All subjects had attended a family practice clinic and controls were women with normal cervical cytology at enrollment. Subjects were asked about tobacco smoke exposure at the workplace and whether they had ever lived with a smoker who had smoked for at least 1 year. The total number of years of exposure, and the relationship of the smoker to the index subject were also asked. For nonsmokers, after adjusting for potential confounders, there was no significant or consistent association between ETS exposure at work or at home and risk of CIN. Analysis by source of ETS exposure showed no association with parents' smoking (OR=0.4, 95% CI=0.1-1.2), a positive association with husbands' smoking (OR=1.5, 95% CI=0.3-6.2) or others’ smoking (OR=1.8, 95% CI=0.4-8.4) after adjustment for age, education, race, number of Pap smears, number of partners and genital warts. We calculated the crude OR for any smoking by husbands (i.e., combine smoking of husband only and of parent) is 2.2 (95% CI=0.9-5.7); the crude OR for any parents' smoking (i.e., combine smoking of parents only and of parent and husband) is 0.9 (95% CI=0.4-2.1) (calculated based on table 5 of Coker et al., 1992). In this study, active smoking was a risk factor irrespective of HPV status; its effect was stronger among women classified as HPV-negative than those classified as HPV-positive.

Data from several other studies on cervical cancer show that husbands of women with cervical cancer are more likely to be smokers than husbands of control women although the effect of husband's smoking generally diminished with adjustment for a woman's smoking. In a study conducted by Buckley et al. (1981), husbands' smoking was associated with an increased risk even after adjustment for wives' smoking habits although its effect diminished. There were too few nonsmoking women to evaluate the role of husbands' smoking in this subgroup. In a study by Hellberg et al (1983, 1986), the effect of husbands' smoking diminished and was no longer statistically significant after adjustment for wives' smoking habits whereas the effect of wives' smoking persisted with adjustment for husbands' smoking. There was, however, a statistically nonsignificant excess of husbands who smoked among nonsmoking wives (Hellberg et al., 1986). In a third study, Zunzunegui et al., (1986) reported an excess of husbands who smoked. Although this excess risk was not adjusted for wives' smoking habits, the authors argued that there was a deficit of smokers among wives and thus wives' smoking is an unlikely explanation for the finding on husbands' smoking.

7.3.2.3 Biomarkers of cervical ETS exposure

In addition to questionnaire-based data, several small studies have been conducted to determine whether there are measurable levels of tobacco smoke constituents in cervical epithelial cells of nonsmokers. Detectable levels of nicotine and cotinine were found in the cervical mucus of nonsmokers; the levels ranged from <1 to about 6 percent of those of active smokers (Sasson et al., 1985; Hellberg et al., 1988; Jones et al., 1991; McCann et al., 1992) (Table 7.10). In three of these studies (Hellberg et al., 1988; Jones et al., 1991; McCann et al., 1992), data were presented separately for nonsmokers exposed to ETS and those with no reported exposure to ETS. In two studies (Hellberg et al., 1988; McCann et al., 1992), levels of nicotine/cotinine in cervical mucus were not distinguishable between nonsmokers with and without ETS exposure whereas in a third study, higher levels of nicotine were found in women with ETS exposure compared to those with no reported exposure (Jones et al., 1991) (Table 7.10). None of the studies on cervical cancer has examined risk of cancer in relation to presence or absence of nicotine/cotinine in the cervical mucus, but this evidence from cross-sectional clinical studies supports the hypothesis that cervical exposure to tobacco constituents occurs from exposure to tobacco smoke.

The presence of carcinogen-DNA adducts in human tissues has been used as evidence of smoking-induced DNA damage. Using 32P-postlabelling techniques, a linear relationship between cigarette consumption and levels of aromatic DNA adducts has been demonstrated in human bronchial epithelium (Phillips et al., 1990a) and other tobacco-related sites, including the cervical epithelium (Phillips et al., 1990b; Cuzick et al., 1990). Recently, Simons et al. (1993) measured levels of DNA adducts in cervical biopsies of 39 women admitted for hysterectomy for benign disease or colposcopy. In this group, 18 were smokers (11 current, 4 exsmokers who stopped in the last 6 months, 3 longer-term exsmokers) and 21 had never smoked. Of the nonsmokers, 75% (n=16) reported exposure to ETS at work or in the home. Urinary cotinine/creatinine levels were also available on these subjects; a ratio of 0.06 or greater was used to indicate active smoking in the previous 24-48 hours.

The median DNA adduct level (per 108 nucleotides) was 4.62 in self-reported smokers which was significantly higher than that in self-reported nonsmokers (3.47). Seven self-reported nonsmokers showed a urinary cotinine/creatinine ratio of 0.06 or greater and they were reclassified as smokers (n=25). The median DNA adduct level in self-reported and reclassified smokers was 4.45 compared to 3.52 (p=0.07) in confirmed nonsmokers (i.e., urinary cotinine/creatinine <0.06). The presence of adducts in cervical epithelium and the correlation with smoking habit strongly suggests that the adducts are a consequence of exposure to tobacco constituents. These results provide direct biochemical evidence that potentially carcinogenic agents may affect the DNA of cervical epithelial cells. It is notable that all the women in the study had detectable proportions of DNA adducts regardless of their smoking status. The relatively high DNA adduct levels in nonsmokers may reflect exposure to ETS, reported by 75% of nonsmokers in this study. Future studies on DNA adduct levels by self-reported exposure to ETS in nonsmokers are needed to confirm these suggestive results.

7.3.2.4 Summary

There is supportive evidence from epidemiological and biochemical studies implicating a role for ETS exposure in the etiology of cervical cancer in nonsmokers. A positive, but nonsignificant association was reported in one cohort study (Hirayama, 1981) and a significant, positive association was observed in two (Sandler et al., 1985a; Slattery et al., 1989) of three case-control studies. In the third case-control study, conducted by Coker et al. (1992), spousal ETS was associated with an increased the risk of cervical cancer/intraepithelial neoplasia in nonsmokers although the result was of borderline statistical significance. Any exposure to ETS (i.e., parents and spouses combined) was not a risk factor in the study by Coker et al. (1992); this finding is not too surprising since risk of cervical cancer appears to be most affected by current tobacco use. Levels of nicotine in the cervical mucus of nonsmokers exposed to ETS were reported to be higher than those with no exposure in one of three biochemical studies. Demonstration of detectable levels of nicotine and cotinine in cervical mucus of nonsmokers suggests that constituents of cigarette smoke may reach more distant sites such as the cervix and play a direct mutagenic role in the etiology of cervical cancer. In addition, the presence of DNA adduct levels in the cervical epithelium of nonsmokers supports the hypothesis that carcinogenic constituents of tobacco smoke may adversely effect the cervical epithelium.

Little is known about the transport of nicotine and cotinine throughout the body and about its metabolism in distant organ sites. Mutagenicity of semen due to smoking is plausible and direct cervical contact with semen of smoking partners may represent another source of exposure to tobacco constituents. It is important to confirm these findings, to determine the importance of recent exposure to ETS (i.e., within recent 5 years) versus lifetime exposure to ETS, and to determine the effect of exposure from spouses versus other household members or coworkers. It is also important to evaluate the effect of passive smoking by stage of cervical cancer (e.g., invasive and pre-invasive), and by history of potential confounding factors, including HPV infection. Measurement of levels of cotinine/nicotine in cervical mucus as well as DNA adducts in cervical epithelium of nonsmokers will complement the epidemiological findings from questionnaires, although such measurements may not be available for cervical cancer cases who are enrolled in studies after surgical treatment for their cancers.

7.3.3.1 Active smoking and bladder cancer

Active smoking is firmly established as a cause of bladder cancer; the relative risks for active smoking ranged from 2 to 10 in different studies (IARC, 1986). The estimated attributable risk for bladder cancer due to smoking is 47% in men and 38% in women (Shopland et al., 1991). The range in relative risk estimates has been explained partly by the different types of tobacco smoked in different countries and the differences in carcinogenicity of tobacco types. Black tobacco products, commonly smoked in countries such as Italy and Argentina, are associated with higher risks of bladder cancer (Vineis et al., 1984; Iscovich et al., 1987) than blond tobacco products, smoked in the US and Canada (Hartge et al., 1990; Burch et al., 1989). Black tobacco, compared to blond tobacco, contains higher concentrations of various aromatic amines, including 4-aminobiphenyl, an established bladder carcinogen (Patrianakos and Hoffmann, 1979; IARC, 1972).

7.3.3.2 ETS and bladder cancer

Risk of bladder cancer in nonsmokers in relation to ETS exposure was evaluated in two studies (Table 7.11).

The first study was conducted by Kabat et al. (1986) as part of a large on-going case-control study of smoking and cancer. Between 1976 and 1983, a total of 948 bladder cancer cases (751 male and 197 female) were interviewed, 152 of whom (76 male and 76 female) were lifetime nonsmokers (i.e., smoked less than 1 cigarette, cigar, or pipe per day for 1 year). Hospital controls, who were also lifetime nonsmokers, were matched to each case on age, sex, race, hospital, and year of interview. There were a total of 492 nonsmoking controls (238 male and 254 female). Questions on ETS exposure were added to the questionnaire in 1979; this information was available on only 40 of 152 cases and 75 of 492 controls interviewed. Questions were asked regarding exposure to ETS inside the home, represented by spouses' smoking, and exposure outside of the home, including exposure at work or in transportation. Results were presented in terms of hours of ETS exposure per week.

The findings on the relationship between ETS exposure and bladder cancer were inconsistent by gender and by source of exposure. For nonsmoking males, there was a nonsignificant increased risk of bladder cancer associated with ETS exposure at home but not at work, whereas among nonsmoking females, a nonsignificant increased risk was observed for ETS exposure at work but not at home (Table 7.11). This study has several limitations, however. The most serious ones include the small sample size of nonsmokers, and that controls were selected from among hospital patients. Although controls were diagnosed with presumably non-tobacco-related diseases (specific diagnoses were not specified), many malignant and non-malignant diseases are causally related to tobacco smoke. Hence, it is quite conceivable that hospital controls may be more likely to be exposed to ETS than the general population.

A second study was conducted by Burch et al. (1989) in Canada between 1979 and 1982. This study included 826 histologically-confirmed bladder cancers and 792 randomly selected controls (Table 7.11). Of these, 142 cases and 217 controls were nonsmokers (defined as having smoked fewer than 185 cigarettes in total). Subjects were asked about their exposure to the tobacco smoke of others at home and at work. For all subjects and for nonsmokers, there was no association between risk of bladder cancer and ETS exposure at home or at work. The authors suggested that because of the modest risk associated with active smoking (RR=2.7) in this study, any association between ETS and bladder cancer in nonsmokers may be too weak to be detectable in questionnaire-based epidemiologic studies.

7.3.3.3 Biomarkers of exposure to bladder carcinogens from ETS exposure

Aside from questionnaire-based epidemiologic studies, some data are available from biochemical measurement studies which evaluated the effect of ETS and risk of bladder cancer in nonsmokers. These studies measured hemoglobin (Hb) adducts of 4- or 3-aminobiphenyl (4-ABP or 3-ABP) which are formed over the 120-day lifespan of the erythrocyte and therefore may serve as dosimeters of average exposure over the previous four months. As mentioned above, the aromatic 4-ABP is a potent human bladder carcinogen (IARC, 1972).

In one study, concentrations of adducts of 4- and 3-ABP were measured in 57 nonsmokers. Subjects who reported exposure to ETS and had detectable serum cotinine levels showed higher median and mean levels of both adducts than subjects who reported no exposure to ETS and had no detectable cotinine levels. The result was of borderline significance for 4-ABP-Hb and was statistically significant for 3-ABP-Hb (MacClure et al., 1989) (Table 7.12). In a second study, Barstch et al. (1990) extended the investigation of 4-ABP-Hb levels in smokers and nonsmokers by N-acetylation phenotype, a marker of susceptibility for bladder cancer (Table 7.13). It has been established that at the same level of exposure to active smoking and other exposures to xenobiotics, slow acetylators are at higher risk of bladder cancer than fast acetylators (Cartwright et al., 1982; Vineis et al., 1990). Among nonsmokers in this study, those with ETS exposure showed higher levels of ABP adducts than those with no ETS exposure. However, the relative increase in ABP adducts differed for 'slow' and 'fast' acetylators. Among nonsmokers with no ETS exposure, the ABP levels were at least two times higher among 'slow' than 'fast' acetylators. However, the ABP levels among nonsmokers with ETS exposure were comparable for 'fast' and 'slow' acetylators. Thus, the increase in ABP levels in relation to ETS exposure was more apparent for 'fast' than 'slow' acetylators (Table 7.13). It is of note that in both studies, nonsmokers showed levels of hemoglobin adducts of 4-ABP that were 28-35% of those of smokers, and the levels of 4-ABP were somewhat higher in nonsmokers exposed to ETS than those not exposed. Levels of 4-ABP were 7% higher in nonsmokers exposed to ETS compared to nonsmokers not exposed in one study (MacClure et al., 1989). In a second study, 4-ABP levels were 14% higher in nonsmokers exposed than nonsmokers not exposed among 'slow' acetylators, and were almost two times higher among exposed 'fast' acetylators compared to non-exposed 'fast acetylators'.

Future studies need to confirm and better characterize the relationship between levels of hemoglobin adducts in nonsmokers and their exposure to ETS by acetylator status.

7.3.3.4 Summary

In summary, the evidence from questionnaire-based epidemiologic studies of ETS and bladder cancer is inadequate. There have been two case-control studies to date, both showed no significant increased risk associated with ETS exposure. These studies, however, had serious limitations including small samples sizes and crude assessment of exposure to ETS. On the other hand, the evidence from two biochemical studies are suggestive. In both studies, nonsmokers exposed to ETS showed higher levels of hemoglobin adducts of an established bladder carcinogen than nonsmokers not exposed to ETS, providing supporting evidence that nonsmokers exposed to ETS may be at increased risk of bladder cancer.