[Surgeon General’s Smoking Report,  health reports Continued from Part  94 ]

2004 The Health Consequences of Smoking: A Report of the Surgeon General.  Office of the Surgeon General (US); Office on Smoking and Health (US). Atlanta (GA): Centers for Disease Control and Prevention (US);  Forty years of reports. Cancers of stomach, uterine cervix, pancreas, kidney, acute myeloid leukemia, pneumonia, aneurysm, cataract, periodontitis. Conclusion “smoking generally diminishes the  health of smokers.”

https://www.ncbi.nlm.nih.gov/books/NBK44698/

Part 95 The Health Consequences of Smoking: A Report of the Surgeon General. 2004. Cancer. Cervical, ovarian, endometrial, stomach, colorectal, prostate, leukemia, liver, brain.

2 Cancer

Cervical Cancer

Cancer of the cervix is one of the leading causes of morbidity and mortality in women throughout the world. In the United States, rates have declined substantially during the past 50 years, reflecting in part a success of screening. In 2003, an estimated 12,200 new cases of cervical cancer were diagnosed, and an estimated 4,100 women were expected to die from this cancer (ACS 2003). From 1996–2000, the incidence in black women (7.0 per 100,000) was higher than in white women (4.7 per 100,000) (Ries et al. 2003). As cervical cancer screening with Papanicolaou smears has become more widespread, the diagnosis of carcinoma in situ has become far more common, and fortunately, invasive carcinoma of the cervix less common.

Cervical cancer is closely linked to sexual behaviors and sexually transmitted infections with human papilloma virus (HPV) (Bosch et al. 2002). In fact, HPV is now considered to be a necessary cause of cervical cancer. Women who begin having sex at a younger age, who have had many sexual partners, or whose partners have had many partners are at a higher risk of developing this disease, likely through increased risk for HPV infection. Against this background, the principal epidemiologic challenges have been to separate the effects of cigarette smoking from the risk factor profile associated with low socioeconomic status, which currently is strongly associated with smoking, and to explore possible causal pathways by which smoking may act with HPV in causing cervical cancer.

Conclusions of Previous Surgeon General’s Reports

The topic of smoking and cancer of the uterine cervix was first reviewed in the 1982 Surgeon General’s report (USDHHS 1982), which concluded that further research was necessary to define whether there was an association between cigarette smoking and cervical cancer. Subsequently, the 1989 report (USDHHS 1989) reviewed more than 15 epidemiologic studies consistently showing an increased risk for cervical cancer in cigarette smokers. Supportive biochemical studies that have detected products of cigarette smoke in cervical mucosa provided a plausible biologic basis for the relationship between cigarette smoking and cervical cancer (USDHHS 1989).

The 1990 report (USDHHS 1990) examined changes in cervical cancer risks after smoking cessation. In the studies that were reviewed, the RR of cervical cancer among current smokers compared with persons who had never smoked ranged from 1.0 to 5.0. After the first year of not smoking, former smokers had lower cervical cancer risks than continuing smokers. The report concluded that the observed diminution in risk after cessation lends support to the hypothesis that smoking is a contributing cause of cervical cancer.

The 2001 report on women and smoking (USDHHS 2001) concluded that smoking has consistently been associated with an increased risk of cervical cancer. It reviewed a large number of case-control studies of invasive cervical cancer and cervical intraepithelial neoplasia, finding smoking to be associated with increased risk in most. However, the report also concluded that the extent to which this association is independent of HPV infection is uncertain. The 2001 report also noted substantial advances in understanding the biology of cervical cancer, notably the role of HPV in carcinogenesis.

Biologic Basis

During the two decades that the Surgeon General’s reports have considered smoking and cervical cancer, there have been substantial advances in understanding the role of HPV in causing this malignancy. In almost all cases, HPV DNA can be identified in the tissue, implying that HPV is necessary to cause cervical cancer (Bosch et al. 1995; Walboomers et al. 1999). In the current pathogenetic model for cervical cancer, smoking might act to increase the rate at which malignancy develops in women with persistent infection or possibly to increase the risk for persistent infection.

A range of evidence supports a possible causal association between cigarette smoking and cervical cancer. Cervical mucous in smokers is mutagenic (Holly et al. 1986) and contains nicotine (McCann et al. 1992) and the carcinogen NNK (Prokopcyzk et al. 1997). DNA adducts reflecting damage to DNA by tobacco products were significantly higher in cervical biopsies of smokers compared with nonsmokers (Phillips and Shé 1994). The adducts detected were consistent with tobacco smoking based on comparisons with tobacco-related adducts found in other tissues. Similar results were reported by the same investigators in a second sample of women undergoing a colposcopy or hysterectomy (Simons et al. 1994). Further studies of DNA adduct formation in normal and HPV-16 immortalized human epithelial cervical cells in cultures show that HPV-16 immortalized cells had significantly greater levels of adducts than did normal cells (Melikian et al. 1999). In vitro model systems also have been used to show that smoking may have an effect on the progression of HPV-initiated carcinogenesis of cervical cancer (Nakao et al. 1996).

Epidemiologic Evidence

As an understanding of the role of HPV in causing cervical cancer has advanced, the approach taken in epidemiologic investigations of smoking has also evolved. In the earliest studies, which antedated any consideration of HPV, smoking was treated as a potential independent risk factor, and possible confounding by indicators of sexual behavior was considered (Winkelstein 1977). As the role of HPV was recognized, investigators attempted to control for HPV by introducing indicators for HPV positivity into risk models or stratifying by HPV status. In these studies, the HPV-negative women with cervical cancer probably included many HPV-positive women incorrectly classified by the early, insensitive-HPV tests. We now have evidence from prospective cohort studies that appropriately reflect the recurring presence of HPV in causing cervical cancer: studies that follow HPV-positive women and compare incidence of cervical cancer precursors in smokers and nonsmokers (Moscicki et al. 2001; Castle et al. 2002).

The Surgeon General’s report on women and smoking (USDHHS 2001) summarized studies of smoking and cervical cancer as well as studies of smoking and intraepithelial neoplasia. An excess risk of cervical cancer among cigarette smokers has been observed in a number of case-control studies, particularly those that controlled for HPV status. However, the extent to which the relationship between smoking and cervical cancer reflects a causal association that is independent of HPV infection was considered uncertain. Studies that did not adjust for HPV status show a RR of approximately 2.0 for current smokers compared with women who never smoked. The risk of cervical cancer increases with the duration of smoking. In two studies of women with a history of smoking for more than 20 years, one found a RR of 4.0 (Peters et al. 1986) and the other a RR of 2.8 (Daling et al. 1996) when compared with women who had never smoked. As summarized in the report on women and smoking (USDHHS 2001), the association between smoking and cervical cancer is seen for both invasive cervical cancer and for precursor conditions, including carcinoma in situ and cervical dysplasia (also known as squamous intraepithelial neoplasia). For premalignant lesions, former smokers have a consistently lower RR than current smokers.

The evidence on cervical canc

er has only recently included studies that took into account HPV status by stratifying on infection status. Early studies in Latin America did not find an independent effect for smoking after controlling for HPV. Several studies that considered HPV status reported that smoking was not associated with a risk of cervical cancer among HPV-positive women (Bosch et al. 1992; Muñoz et al. 1993; Eluf-Neto et al. 1994). In Latin American countries, women generally smoke small numbers of cigarettes daily, however, and findings are different in other countries.

Among women who tested positive for HPV, two studies found smoking to be a risk factor in both HPV-positive and HPV-negative women. In a population-based, case-control study of invasive cervical cancer in western Washington state, Daling and colleagues (1996) found women with cervical cancer were more likely to be current smokers at diagnosis than population controls (RR = 2.5 [95 percent CI, 1.8–3.4]). The risk associated with smoking was present to a similar extent among women who tested positive and negative for HPV. In a case-control study nested in a population-based cohort consisting of women participating in cytological screening in Sweden, Ylitalo and colleagues (1999) found that after multivariate adjustment, a twofold higher risk was observed among current smokers compared with lifetime nonsmokers (odds ratio [OR] = 1.94 [95 percent CI, 1.32–2.85]), an association apparently confined to women younger than 45 years. Other studies reported since the 2001 report of the Surgeon General also show an association of smoking with cervical neoplasia. In two prospective cohort studies in the United States, smoking was associated with an increased risk in women who were HPV positive on enrollment. Moscicki and colleagues (2001) followed 496 women who were HPV positive over a median of 26 months. Daily cigarette smoking was associated with an increased risk for incident low-grade squamous intraepithelial lesion development (relative hazard = 1.67 [95 percent CI, 1.12–2.48]). In a 10-year cohort study of 1,812 Oregon women infected with HPV, women who smoked had an increased risk for high-grade cervical intra-epithelial neoplasia (Castle et al. 2002). Compared with lifetime nonsmokers, the RRs were 2.9 (95 percent CI, 1.4–6.1) for smokers of less than one pack of cigarettes per day, 4.3 (95 percent CI, 2.0–9.3) for one or more packs per day, and 3.9 (95 percent CI, 1.6–6.7) for former smokers (Castle et al. 2002). Two nested case-control studies, one in Costa Rica (Hildesheim et al. 2001) and the other in the United Kingdom (Deacon et al. 2000), had similar findings in HPV-positive women.

Evidence Synthesis

Strong biologic evidence supports a mechanism for direct action of tobacco smoke components on the epithelial cells of the cervix. DNA adducts isolated from cervical cells reflect tobacco exposures among smokers. A large body of epidemiologic evidence supports a positive relationship between smoking and cervical cancer. Smoking has consistently been associated with higher risks of cervical cancer that increase with the duration of smoking and the number of cigarettes smoked per day (USDHHS 2001). Similar associations have been observed for premalignant lesions. Until recently, few studies appropriately considered HPV exposure and infection. HPV is now recognized as a likely contributor to the etiology of most cases and that the risk of smoking is most appropriately assessed in HPV-positive women. The most recent studies consistently show that smoking is associated with an increased risk among HPV-positive women. The increased risk is of a moderate strength and not likely to be explained by confounding by sexual behavior, as all women were HPV-positive in these analyses. Dose-response relationships were also demonstrated. Finally, in 2002, IARC concluded that there is now sufficient evidence for a causal association between cigarette smoking and cancer of the uterine cervix (IARC 2002).

Conclusion

1. The evidence is sufficient to infer a causal relationship between smoking and cervical cancer.

Implication

Further study to refine epidemiologic and mechanistic understanding of the independent association between smoking and HPV infection will clarify the causal association between smoking and cervical cancer.

Ovarian Cancer

Ovarian cancer is a leading cause of cancer mortality among women. In 2003, an estimated 25,400 new cases and 14,300 deaths attributed to this cancer were expected to occur. It ranks second among gynecologic cancers, and accounts for nearly 4 percent of all cancers among women (ACS 2003). From 1900–1970, ovarian cancer rates increased, perhaps reflecting changes in childbirth toward smaller families. Incidence and mortality have decreased slightly since 1970, probably reflecting the use of oral contraceptives, a known protective factor against ovarian cancer (Hankinson et al. 1992; McKean-Cowdin et al. 2000).

Conclusions of Previous Surgeon General’s Reports

Ovarian cancer was first addressed in the 2001 Surgeon General’s report on women and smoking (USDHHS 2001), which noted that smoking is probably not related to ovarian cancer.

Biologic Basis

A broad range of possible biologic mechanisms could lead to an effect of smoking on ovarian cancer risks, reflecting the effects of smoking on ovarian tissue and possibly female hormones. Evidence supports the possibility that cigarette smoke products and their metabolites act directly on tissue with estrogen receptors. Smoking may also influence risks by modifying hormone levels (see the section on “Breast Cancer” later in this chapter for a review of the hormonal effects of cigarette smoking). Metabolic products of tobacco smoke can be found in ovarian follicular fluid as can indicators of oxidative stress (Hellberg and Nilsson 1988; USDHHS 1990; Paszkowski et al. 2002). Alkaloids in cigarette smoke have been shown to inhibit corpus lutea progesterone synthesis (Gocze et al. 1996). In a model with primary granulosa cells, the alkaloids and smoke extract decreased DNA production, suggesting a cytotoxic effect. This wide range of potential effects of tobacco smoke could potentially influence the risks of ovarian cancer either directly or indirectly.

Epidemiologic Evidence

The available epidemiologic evidence is not consistent with regard to the strength of an association between smoking and ovarian cancer, or with regard to the temporal changes in risks following smoking cessation. Although some case-control studies have not distinguished current smokers from former smokers (Polychronopoulou et al. 1993; Purdie et al. 1995), others that have separately evaluated current and former smokers observed few differences between these two groups in the risk of ovarian cancer (Franks et al. 1987; Stockwell and Lyman 1987).

A recent study of the relationship between smoking and histologic subtypes of ovarian cancer found a RR of 2.9 (95 percent CI, 1.7–4.9) for mucinous epithelial tumors when comparing current smokers with those who had never smoked (Marchbanks et al. 2000). These data come from a population-based, case-control study that included 447 cases of ovarian cancer and 3,868 controls. This elevated risk was evident regardless of the age at smoking initiation, although the risk increased slightly as the cumulative pack-years of smoking increased. Similar patterns of risk were not observed among serous, endometrioid, or other histologic types. In a population-based, case-control study conducted in Australia, Green and colleagues (2001) observed a similar relationship. In an analysis of 794 cases and 855 controls, the histologic subtype of ovarian cancer most strongly related to cigarette smoking was the mucinous subtype. For current smokers, the RR was 3.1 (95 percent CI, 1.8–5.4) compared with women who had never smoked, and the risk of mucinous ovarian cancer increased with the maximum number of cigarettes smoked per day. For nonmucinous tumors, the RR was 1.5 (95 percent CI, 1.1–2.1) for smokers compared with nonsmokers.

Evidence Synthesis

Data on the relationship between cigarette smoking and ovarian cancer remain inconclusive. Evidence for patterns of risks with the duration of smoking and time since quitting is limited. Histologic subtypes of ovarian cancer appear to have distinct etiologic factors. Consistent findings suggest that a relationship to cigarette smoking for the mucinous subtype of ovarian cancer is plausible (Marchbanks et al. 2000; Green et al. 2001).

Conclusion

1. The evidence is inadequate to infer the presence or absence of a causal relationship between smoking and ovarian cancer.

Implication

Further research is needed to evaluate risks by histologic subtypes, to evaluate duration of smoking and risk, and to determine the time course of risk following smoking cessation.

Endometrial Cancer

Cancer of the endometrium (uterine corpus) is now the most commonly occurring gynecologic malignancy in women. In 2003, an estimated 40,100 new cases and 6,800 deaths were expected to occur from endometrial cancer (ACS 2003). Incidence rates are higher in white women (14.0 per 100,000) than in black women (10.0 per 100,000), but mortality rates are nearly twice as high for black women (Ries et al. 2003).

Endometrial cancer risks are predominantly determined by various hormonal risk factors: exposures to estrogens from estrogen replacement therapy after menopause, the use of tamoxifen, early menarche or late menopause, nulliparity, and a failure to ovulate (except while taking oral contraceptives). Obesity is also associated with increased risk. Pregnancy and the use of combination oral contraceptive pills (which include both estrogen and progesterone) are each protective against endometrial cancer (Grady and Ernster 1996).

Because of the strong dependence of endometrial cancer risk on exposure to estrogens, separating direct and indirect causal pathways for the effect of smoking on ovarian cancer risk has been difficult. Women who smoke are more likely to be lean and to enter menopause earlier than nonsmokers (Willett et al. 1983). They are thus more likely to take estrogen therapy after menopause and to have more years of estrogen exposure (Pike et al. 1998). Separating causal paths involving smoking from those involving hormonal factors has consequently been complicated.

Conclusions of Previous Surgeon General’s Reports

The inverse relationship between cigarette smoking and the risk of endometrial cancer was first noted in the 1989 Surgeon General’s report (USDHHS 1989). Endometrial cancer is less frequent in women who smoke cigarettes. The 2001 Surgeon General’s report on women and smoking (USDHHS 2001) updated this conclusion by noting that current smoking is associated with a reduced risk of endometrial cancer, although the effect is probably limited to postmenopausal women. The risk of endometrial cancer in former smokers generally appears more similar to that in women who have never smoked.

Biologic Basis

As reviewed in the section on “Breast Cancer” later in this chapter, several lines of evidence support a biologic pathway for cigarette smoking in influencing hormone levels from exogenous estrogen and the risk of hormone-related cancers. Such potential pathways include an altered metabolism as well as a lower production of estrogens because of lower adiposity.

Epidemiologic Evidence

More recent studies continue to show a reduced risk for endometrial cancer in smokers compared with nonsmokers. In a cohort study of participants in the Canadian Mammography Screening Trial, risk was reduced in current smokers compared with lifetime nonsmokers, but only among those smoking 20 or more cigarettes per day (hazard ratio = 0.62 [95 percent CI, 0.42–0.92]) (Terry et al. 2002). Case-control studies in Wisconsin (Newcomer et al. 2001), Washington state (Littman et al. 2001), and Sweden (Weiderpass and Baron 2001) also provide evidence of a reduced risk in smokers compared with nonsmokers (Table 2.18).

Table 2.18

Studies on the association between smoking and the risk of endometrial cancer. 

Evidence Synthesis

A consistent association between smoking and a lower risk of endometrial cancer has been found. The biologic basis for this association is consistent with the antiestrogenic effect attributed to smoking.

Conclusion

1. The evidence is sufficient to infer that current smoking reduces the risk of endometrial cancer in postmenopausal women.

Implication

Because smoking has numerous adverse health effects as summarized in this report, the modest reduction in the risk of endometrial cancer associated with smoking is far outweighed by the increase in other causes of smoking-related morbidity and mortality.

Stomach Cancer

Despite a major decline in the incidence of stomach cancer in industrialized countries across the last century, gastric carcinoma remains the second most common fatal cancer worldwide (Pisani et al. 1999). An estimated 22,400 new cases and 12,100 deaths from cancer of the stomach were expected to occur in the United States in 2003 (ACS 2003).

Incidence and death rates for stomach cancer vary by race, gender, and ethnicity. Incidence is approximately twice as high among men as among women and higher among nonwhites than whites. A substantial variation of incidence is evident among both men and women, respectively, across various racial and ethnic groups: Asian/Pacific Islanders (23.0 and 12.8), blacks (19.9 and 9.9), Hispanics (18.1 and 10.0), American Indians/Alaska Natives (14.4 and 8.3), and white non-Hispanics (10.0 and 4.3). In the United States, the median survival of persons with stomach cancer is less than one year after diagnosis, although the relative five-year survival rate has increased slightly from 15.1 percent for patients diagnosed in 1975 to 22.5 percent for patients diagnosed in 1992 (Ries et al. 2000a, 2003).

Internationally, death rates from stomach cancer vary nearly 100-fold across countries (IARC 2003). Stomach cancer is the most common malignancy in China and in parts of eastern Asia and Latin America (Parkin et al. 1999; Pisani et al. 1999). Mortality rates have been decreasing worldwide but are as high as 50 per 100,000 among men and 26 per 100,000 among women in the highest risk countries (IARC 2003).

Assessments of the independent contribution of cigarette smoking to the development of stomach cancer are complicated by two factors. First, the background occurrence of stomach cancer decreased globally during much of the twentieth century for reasons unrelated to changes in cigarette smoking. This decline is exemplified by the falling mortality rate from stomach cancer in the United States since 1930, when cause-specific national mortality statistics first became available (Figure 2.6) (Greenlee et al. 2000). The age-adjusted mortality rate (per 100,000) decreased 85 percent in men and 90 percent in women between 1930 and 1997. Figure 2.6 also shows the increase in per capita use of manufactured cigarettes that began in the early 1900s and persisted through 1963 (Giovino et al. 1994), coinciding with much of the decrease in stomach cancer mortality. The main factors proposed to account for the decline in stomach cancer are the introduction of refrigeration (with the resultant increased availability of fresh fruits and vegetables and reduced consumption of salted, smoked, and pickled foods), improved sanitation, and the introduction of antibiotic therapy (reducing chronic Helicobacter pylori(H. pylori) infections) (Nomura 1996). It has been challenging to identify the contribution to stomach cancer risk from cigarette smoking in the context of large temporal changes in other apparently important risk factors.

Figure 2.6

Stomach cancer death rates stratified by gender and per capita number of cigarettes smoked in the United States, 1930–1994. Sources: Centers for Disease Control and Prevention, National Center for Health Statistics, U.S. Mortality Volumes 1930– (more…)

A second challenge in determining whether cigarette smoking causes stomach cancer is that the gastric cancers at different subsites appear to differ etiologically, yet are combined in most epidemiologic studies. Subsites of stomach cancer usually are not considered in mortality studies, because death certificates seldom record the histology or location of the tumor within the stomach. The predominant type of stomach cancer observed in incidence registries in the United States and Europe has changed over time, particularly among men. The incidence of cancers of the gastric cardia subsite, occurring near the junction of the esophagus with the stomach, increased by 4.3 percent annually among men in United States SEER areas between 1976 and 1987 (Devesa and Fraumeni 1999). A similar increase in gastric cardia cancers has been observed in Europe (Golematis et al. 1990; Craanen et al. 1992; Botterweck et al. 2000), at the same time that the incidence of cancers of the gastric antrum, corpus, or fundus (termed noncardia cancers) has been decreasing worldwide. The decline in noncardia cancers accounts for most of the global decline in stomach cancer. As a consequence of these opposing trends, tumors of the gastric cardia now compose about one-third of all stomach cancers among white men in the United States (Blot et al. 1991).

Conclusions of Previous Surgeon General’s Reports

Stomach cancer has not been classified among the diseases definitely caused by tobacco smoking by the Surgeon General (USDHEW 1964, 1974; USDHHS 1982, 1989a) or IARC until the most recent monographs (IARC 2002). However, the evidence supporting a causal relationship has become stronger over time. Key conclusions from previous Surgeon General’s reports are presented as follows by year:

No relationship has been established between tobacco use and stomach cancer (USDHEW 1964, p. 229).

No firm relationship between stomach cancer and cigarette smoking has been established (USDHEW 1974, p. 55).

In epidemiological studies, an association between cigarette smoking and stomach cancer has been noted. The association is small in comparison with that noted for smoking and some other cancers (USDHHS 1982, p. 22).

Evidence from prospective and retrospective studies available more recently has shown a small but consistent increase in mortality ratios [for stomach cancer], averaging approximately 1.5 for smokers compared with nonsmokers. Dose-response relationships have been demonstrated for the number of cigarettes smoked per day (USDHHS 1989, p. 57).

Tobacco has been associated with stomach cancer, but whether this association is causal remains unclear (USDHHS 1990, p. 176).

Biologic Basis

More than 90 percent of stomach cancers diagnosed in the United States are adenocarcinomas, the remainder being predominantly non-Hodgkin’s lymphomas or leiomyosarcomas (Rotterdam 1989; Fuchs and Mayer 1995). Gastric adenocarcinoma is further subdivided into two histopathologic categories: an intestinal or glandular subtype (in which the cells resemble intestinal columnar epithelium and form gland-like, tubular structures) and a diffuse form (characterized by poorly cohesive tumor cells that infiltrate and thicken the stomach wall without forming a discrete mass) (Fuchs and Mayer 1995; Nomura 1996). The intestinal subtype is the predominant noncardia cancer in regions where the risk for noncardia cancer is high and where the intestinal subtype accounts for most of the excess risk (Correa 1992). Clinical differences between intestinal and diffuse gastric cancers are that the former occur at older ages, more frequently in the distal stomach, and are usually preceded by several decades of chronic gastritis, inflammation, and premalignant abnormalities (Correa 1992; Fuchs and Mayer 1995).

Cigarette smoking was associated with more severe premalignant gastric abnormalities in a population-based study that performed gastroscopic examinations on approximately 3,000 residents of Linqu County, China, in 1989 and 1990 (Kneller et al. 1992). This region has one of the highest rates of gastric cancers in the world (mostly of the intestinal subtype). Smokers were more likely than nonsmokers in the study to have been diagnosed with intestinal metaplasia and/or dysplasia. Nonsmokers were more likely than smokers to have the less severe superficial gastritis and/or chronic atrophic gastritis. The risk for dysplasia increased with the number of cigarettes smoked per day and years of smoking (Kneller et al. 1992). The authors attributed virtually all of the 55 percent higher prevalence of gastric dysplasia in men than in women to the higher smoking prevalence in men (80 percent) versus women (5 percent). A second endoscopic examination of persons in this study in 1994 demonstrated longitudinally that persons with more severe baseline lesions were more likely to experience progression to dysplasia or a gastric cancer (You et al. 2000).

Although certain somatic mutations are frequently observed in genetic studies of gastric adenocarcinomas, there is as yet no well-defined molecular model of tumorigenesis (Powell 1998), and specific genetic changes have not been studied in relation to cigarette smoking. Somatic mutations of the p53 tumor suppressor gene are detected in 60 percent of gastric adenocarcinomas of both histologic types (Powell 1998). Mutations in p53 are most often observed in the advanced stages of gastric dysplasia rather than as an early stage in carcinogenesis. Other genetic changes associated with gastric adenocarcinomas include deletions and amplifications of the gene for transforming the growth factor beta type II receptor, the deleted DCC gene in colon cancer, and the candidate tumor suppressor genes DPC4 and madd(Tahara 1995; Powell 1998). A subset of gastric tumors also displays microsatellite instability (Gong et al. 1999) similar to that seen in a subset of colon cancers from hereditary nonpolyposis coli families predisposed to various malignancies. Molecular changes that may be unique to the diffuse type of gastric cancers include the reduction or loss of cadherins and catenins and amplification of K-sam genes. Unique to the intestinal type are K-rasmutations, erbB-2 gene amplification, loss of heterozygosity and mutations of the APC gene, and loss of heterozygosity of the bcl-2 and DCC genes (Gong et al. 1999).

Nicotine and other components of cigarette smoke affect several aspects of gastric physiology (reviewed in detail in the section on “Peptic Ulcer Disease” in Chapter 6). Short-term effects of smoking include increased reflux of duodenal contents into the stomach and mouth, decreased secretion of pancreatic bicarbonate, decreased production of gastric mucus and cytoprotective prostaglandins, and perhaps the increased production of free radicals and release of vasopressin, a potent vasoconstrictor (Endoh and Leung 1994; Eastwood 1997).

Studies have begun to examine whether cigarette smoking influences other environmental risk factors for stomach cancer, particularly H. pyloriinfections (Ley and Parsonnet 2000). Properly designed studies are needed to sort out the causal pathways for stomach cancer and smoking and H. pyloriinfections. Smoking, for example, might act to increase the risk for infection or to synergistically modify the carcinogenic processes associated with infections. The prevalence of a H. pylori infection is reported to be higher among smokers than among lifetime nonsmokers in some cross-sectional studies (Graham et al. 1991; Bateson 1993; Brenner et al. 1997; Goh 1997; Murray et al. 1997; Lin et al. 1998; Phull et al. 1998; Collett et al. 1999), but not in all of them (Maxton et al. 1990; Lindell et al. 1991; Battaglia et al. 1993; EUROGAST Study Group 1993; Tsugane et al. 1994; Shinchi et al. 1997; Russo et al. 1999; Ogihara et al. 2000). Several studies also report that the eradication of an H. pylori infection with antibiotics is more difficult in smokers than in non-smokers (Cutler and Schubert 1993; O’Connor et al. 1995; Goddard and Spiller 1996; Bardhan et al. 1997; Breuer et al. 1997a,b), although at least one study has not found this result (Chan et al. 1997). Thus there is some evidence that cigarette smoking may increase the infectivity of H. pylori or decrease host resistance to the infection, although it remains possible that an H. pylori infection simply is correlated with smoking in some studies.

The combination of an H. pylori infection and cigarette smoking also may be more pathogenic to the gastric mucosa than an H. pylori infection alone. Zaridze and colleagues (2000) observed that among men infected with H. pylori in Russia, those who ever smoked had a twofold higher risk of stomach cancer than nonsmokers (OR = 2.3 [95 percent CI, 1.1–4.7]). This study found no increase in stomach cancer risks among women who smoked or among male smokers uninfected with H. pylori (p value for interaction = 0.07). Another study in Poland found more frequent evidence of intestinal metaplasia in persons infected with H. pylori who smoked cigarettes, consumed vodka, or did both than in those with an H. pylori infection alone (Jedrychowski et al. 1993, 1999).

H. pylori infections may have differing effects on cancers of the gastric cardia than on noncardia cancers (Fox and Wang 2000). Whereas an H. pyloriinfection is an established risk factor for noncardia stomach cancers, some evidence suggests that H. pylori infections actually may be protective against gastric cardia tumors at the gastroesophageal junction (Blaser 1999a,b). Eradication of H. pylori results in increased rates of gastroesophageal reflux, a factor contributing to the pathogenesis of Barrett’s syndrome and esophageal adenocarcinoma (Labenz et al. 1997; Vicari et al. 1998). Persons who carry particular cagA(+) strains of H. pylori experience a marked inflammation of the gastric cardia but have a lower risk of developing adenocarcinoma of either the gastric cardia or the esophagus (Peek et al. 1999; Vaezi et al. 2000).

Compared with nonsmokers, current cigarette smokers have lower plasma and serum concentrations of certain micronutrients, such as beta carotene and ascorbic acid, that may protect against the development of stomach cancer (Smith and Hodges 1987; Stryker et al. 1988; Zondervan et al. 1996). The concentration of these substances in the blood is lower than would be expected from dietary intake (Smith and Hodges 1987; Stryker et al. 1988; Bolton-Smith et al. 1991). It has been proposed that smokers may require a higher dietary intake of certain protective micronutrients than nonsmokers because of a more rapid degradation or excretion of these micronutrients (Stryker et al. 1988; Cross and Halliwell 1993).

Animal models of the carcinogenicity of tobacco smoke to the stomach are limited and largely involve tumors of the rodent forestomach, an organ more analogous to the human esophagus than to the stomach. Specific chemicals found in tobacco smoke and smoke condensate are known to cause cancers of the rodent forestomach when administered orally or by gavage (USDHHS 2000). Substances in cigarette smoke that are listed by the National Toxicology Program as carcinogenic to the rodent forestomach include benz[a]anthracene (mouse: gavage), benzo[a]pyrene (mouse and hamster: gavage), dibenz(a,h)anthracene (mouse: diet), 7H-dibenzo(c,g)carbarole (mouse: gavage), n-nitrosodin-butylamine (mouse and hamster: diet, drinking water, and gavage), and n-nitrosodiethylamine (mouse: diet and gavage) (USDHHS 2000).

Epidemiologic Evidence

This section considers all published studies (in English) that provide separate data on lifetime nonsmokers and current and former cigarette smokers. Where multiple follow-ups have been reported on the same cohort, data from the longest follow-up are presented. Studies were identified by searching the MEDLINE database (from January 1966 to August 2000) using the medical subject headings “tobacco,” “smoking,” “gastric neoplasms,” and “stomach neoplasms,” and by examining references cited in published original and review articles (Trédaniel et al. 1997).

Nine cohort studies (Table 2.19) (Nomura et al. 1990; Kneller et al. 1991; Kato et al. 1992; Tverdal et al. 1993; Doll et al. 1994; McLaughlin et al. 1995a; Engeland et al. 1996; Mizoue et al. 2000; ACS, unpublished data) and 11 case-control studies (Table 2.20) (Correa et al. 1985; Jedrychowski et al. 1986; Boeing et al. 1991; Saha 1991; Agudo et al. 1992; Hansson et al. 1994; Ji et al. 1996; De Stefani et al. 1998; Chow et al. 1999; Inoue et al. 1999; Zaridze et al. 2000) have examined the association between cigarette smoking status and incidence of or death from stomach cancer. Current cigarette smokers consistently have higher incidence or death rates than do lifetime nonsmokers in studies of men (Nomura et al. 1990; Kneller et al. 1991; Tverdal et al. 1993; Doll et al. 1994; McLaughlin et al. 1995a; Engeland et al. 1996; Mizoue et al. 2000; ACS, unpublished data) and men and women combined (Kato et al. 1992); this finding is less consistent in studies of women (Table 2.19) (Tverdal et al. 1993; Engeland et al. 1996; ACS, unpublished data). The average RR estimate among current smokers compared with lifetime nonsmokers across all of the studies in Tables 2.19and 2.20, weighted by the number of cases, is 1.6 (1.7 in men and 1.3 in women). Relative risk estimates above 2.0 are seen in several studies of Japanese (Nomura et al. 1990; Kato et al. 1992; Inoue et al. 1999; Mizoue et al. 2000) and other populations with above average risks of stomach cancer (Kneller et al. 1991; Tverdal et al. 1993; De Stefani et al. 1998).

Table 2.19

Cohort studies on the association between smoking status and the risk of stomach cancer. 

Table 2.20

Case-control studies on the association between smoking status and the risk of stomach cancer. 

Former smokers have lower incidence or death rates for stomach cancer than do continuing smokers in most studies of men (Tables 2.19 and 2.20) (Nomura et al. 1990; Kneller et al. 1991; Tverdal et al. 1993; Doll et al. 1994; McLaughlin et al. 1995a; Ji et al. 1996; De Stefani et al. 1998; Chow et al. 1999; Inoue et al. 1999; Zaridze et al. 2000; ACS, unpublished data), although one study found a higher risk for former smokers in men and women (Kato et al. 1992). The average RR estimate in former smokers across all studies combined is 1.2 (1.2 in men and 1.3 in women).

Among current smokers, most studies document only a small increase in the risk for stomach cancer with an increasing number of cigarettes smoked per day (Tables 2.21 and 2.22) or years of smoking (Table 2.23). Two prospective studies that do show some gradient of an increased risk with a greater number of cigarettes smoked are the reports by Kneller and colleagues (1991) from Norway and McLaughlin and colleagues (1995a) on United States veterans. The tests for a trend presented in Tables 2.21 and 2.22 are taken from the original papers and do not always specify whether lifetime nonsmokers were excluded from the trend calculations. No significant trend is observed with either the number of cigarettes smoked per day (Table 2.22) or number of years of smoking (Table 2.23) in CPS-II (ACS, unpublished data).

Table 2.21

Cohort studies on the association between the number of cigarettes smoked per day and the risk of stomach cancer. 

Table 2.22

Case-control studies on the association between the number of cigarettes smoked per day and the risk of stomach cancer. 

Table 2.23

Cohort studies on the association between current smoking, years of smoking, and the risk of stomach cancer. 

Among former smokers, the risk of stomach cancer consistently decreases below that of continuing smokers with the number of years since cessation (Table 2.24). This trend is clearest in the studies with the largest number of former smokers (De Stefani et al. 1998; ACS, unpublished data). The risk of stomach cancer among former smokers approaches that of lifetime nonsmokers approximately 20 years after quitting.

Table 2.24

Cohort and case-control studies on the association between years since quitting smoking and the risk of stomach cancer. 

The epidemiologic studies that have separated cancers of the gastric cardia from noncardia cancers suggest that cancers at both subsites are associated with cigarette smoking (Table 2.25). Two case-control studies (Kabat et al. 1993; Gammon et al. 1997) report stronger associations between smoking and cancers of the gastric cardia than between smoking and noncardia cancers. However, the evidence relating smoking to specific types of stomach cancer is limited (Nomura 1996), as most studies have not been analyzed by anatomic or histologic subsites.

Table 2.25

Case-control studies on the association between smoking status and the risk of stomach cancer stratified by subsite. 

Evidence Synthesis

A large decrease in stomach cancer incidence and death rates occurred in the United States during the time per capita cigarette smoking increased steeply. The timing of these trends and the continuing decrease in gastric cancer incidence and mortality worldwide suggest that cigarette smoking is not, by itself, a major independent cause of stomach cancer. It nevertheless remains possible that cigarette smoking is an important factor in the pathogenesis of both cardia and noncardia stomach cancers.

Many large, well-conducted epidemiologic studies consistently report higher incidence or death rates for stomach cancer among current cigarette smokers than among lifetime nonsmokers. Studies that distinguish between cancers of the gastric cardia and those elsewhere in the stomach generally find that smoking is associated with both sites. Persons who stop smoking have a lower risk of stomach cancer than those who continue. The risk among former smokers diverges progressively away from that of continuing smokers and toward that of lifetime nonsmokers as time elapses after cessation. Among current smokers, the risk of stomach cancer is not strongly associated with either years of smoking or the number of cigarettes smoked per day. In 2002, IARC concluded that there is now sufficient evidence for a causal association between cigarette smoking and cancer of the stomach (IARC 2002).

Cigarette smoking may increase the infectivity or add to the pathogenicity of H. pylori, a known cause of noncardia stomach cancer. The prevalence of Helicobacter infections is inconsistently reported to be higher among cigarette smokers than among lifetime nonsmokers in some studies. The eradication of H. pylori infections using antibiotics was more difficult in smokers than nonsmokers in several studies. An H. pylori infection in combination with cigarette smoking is associated with more frequent ulcerations (gastric and duodenal combined) (Martin et al. 1989), the progression to metaplasia (Jedrychowski et al. 1993, 1999), and/or gastric cancers (Zaridze et al. 2000) than is anH. pylori infection alone. Cigarette smoking is also thought to deplete the plasma and serum concentrations of certain micronutrients that may protect against Helicobacter infections or gastric neoplasia.

Two important limitations of most of the epidemiologic studies are that few studies have measured infections with H. pylori and cigarette smoking in the same people, and studies have not consistently distinguished between gastric cardia and noncardia cancers. Such information is needed to examine the separate and joint effects of cigarette smoking and an H. pylori infection on the main subtypes of stomach cancer. The interaction between smoking and H. pylori may vary across different subtypes of gastric cancer. Some evidence suggests that H. pylori infections may be negatively associated with cancers of the gastric cardia but positively associated with noncardia gastric cancers (Hansen et al. 1999). The critical exposure for non-cardia cancers may be the combination of an H. pylori infection and cigarette smoking. If so, then conventional dose-response analyses may misclassify the duration or intensity of the relevant exposure by considering one or both of these factors separately.

Conclusions

1. The evidence is sufficient to infer a causal relationship between smoking and gastric cancers.

2. The evidence is suggestive but not sufficient to infer a causal relationship between smoking and noncardia gastric cancers, in particular by modifying the persistence and/or the pathogenicity of Helicobacter pylori infections.

Implications

With inference of a causal association between current and former cigarette smoking and death from gastric cancers, including stomach cancer among the smoking attributable conditions increases the estimated number of deaths caused by smoking by 3,573 in 1990 in the United States, based on CPS-II. The impact of smoking on gastric cancers may be substantially greater in developing countries where the incidence of and mortality from stomach cancer are higher.

Reductions in smoking could help to counteract the increase in cancers of the gastric cardia occurring in the United States and Europe, especially among men. Further research is needed to assess the combined effects of cigarette smoking and an H. pylori infection. Of particular interest is the impact of continued cigarette smoking on the infectivity and pathogenicity of H. pylori, and the relationship of smoking and other factors to cancers of the gastric cardia.

Colorectal Cancer

Together, cancers of the colon and rectum rank as the third most common cancers and cause of cancer deaths among men and women in the United States (ACS 2003). In 2003, an estimated 105,500 cases of cancer of the colon and 42,000 cases of cancer of the rectum were expected to be diagnosed. That same year, 57,100 deaths from both cancers combined were expected to occur (ACS 2003). In the mid-1990s, the lifetime probability of developing colorectal cancer was estimated to be 5.6 percent in the United States (Greenlee et al. 2000).

Worldwide, colorectal cancer incidence and mortality rates vary more than 10-fold among countries; the highest rates occur in western Europe, North America, Australia/New Zealand, and Japan; and the lowest rates occur in countries with developing economies, particularly in Africa and Asia (Parkin et al. 1999; Pisani et al. 1999). Studies of migrants show that, in immigrants moving from countries where the incidence is low to countries where the incidence is high, incidence rates increase within one generation to approximate rates of the new country, suggesting a strong role for environmental causes (Thomas and Karagas 1987; McMichael and Giles 1988).

The average annual age-adjusted population incidence rate of colorectal cancer per 100,000 in the United States from 1996–2000 was 72.4 in black men, 64.1 in white men, 57.2 in Asian/Pacific Islander men, 56.2 in black women, 49.8 in Hispanic men, 46.2 in white women, 38.8 in Asian/Pacific Islander women, 37.5 in American Indian/Alaska Native men, 32.9 in Hispanic women, and 32.6 in American Indian/Alaska Native women (Ries et al. 2003). Incidence rates are consistently higher among men than among women in all racial and ethnic groups (Ries et al. 2003). Colorectal cancer incidence rates increased from 1973 until 1985 and began decreasing steadily in the mid-1980s; mortality rates increased through 1991 and then decreased rapidly through 1997 (Chu et al. 1994; Ries et al. 2000b). The decrease in both incidence and mortality rates has been larger and began earlier in white women than in white men.

The five-year relative survival rate among whites in the United States is approximately 90 percent when colorectal cancers are diagnosed and treated at the localized stage, but falls below 10 percent when they are diagnosed at the distal stage. Fewer than 40 percent of all cases are diagnosed at the localized stage (Ries et al. 2003). A shift toward an earlier stage at diagnosis occurred among white men and women in the United States between 1975 and 1995 (Troisi et al. 1999), and the resulting improvements in survival have been attributed mostly to the earlier removal of localized carcinomas (Chu et al. 1994; Troisi et al. 1999; Ries et al. 2000b).

Colorectal cancer risk factors include physical inactivity, obesity, and perhaps a diet high in saturated and animal fats and low in vegetables and fruits. These risk factors are still under investigation and uncertainty remains, particularly with regard to the specific dietary factors. The risks also increase for persons with a family history of colorectal cancer or polyps. Factors consistently associated with a reduced risk are the use of aspirin and other nonsteroidal anti-inflammatory drugs, and hormone replacement therapy use among women (Potter 1999).

Colorectal cancer was among the causes of mortality assessed in cohort studies. The hypothesis that prolonged cigarette smoking may contribute to colorectal cancer gained support in the mid-1990s when epidemiologic (particularly cohort) studies reported a higher incidence of adenomatous polyps and/or cancer in long-term smokers (Giovannucci et al. 1994a,b). Uncertainty about the reports of this observed association has primarily come from the possibility of uncontrolled confounding by other lifestyle determinants of risk that are still under study (Doll 1996; Giovannucci and Martínez 1996). Giovannucci and Martínez (1996) and Giovannucci (2001)have provided comprehensive reviews of the literature and the methodologic concerns.

Conclusions of Previous Surgeon General’s Reports

Until the 2001 Surgeon General’s report on women and smoking (USDHHS 2001), this series of reports had not considered smoking in relation to cancers of the colon and rectum, and colorectal cancers are not included among the smoking-related cancers by the Centers for Disease Control and Prevention (CDC) (Nelson et al. 1994) or IARC (1986) (Parkin et al. 1994).

Biologic Basis

Most cancers of the colon and rectum are adenocarcinomas (Rosai 1996). These tumors typically develop from clonal expansions of mutated cells through a series of histopathologic stages from single crypt lesions to benign tumors (adenomatous polyp) and then to metastatic carcinomas that take place over a span of 20 to 40 years (Fearon and Vogelstein 1990; Kinzler and Vogelstein 1998). The number and order of genetic and epigenetic changes in tumor suppressor genes (such as APC, p53, and DCC) and oncogenes (such as ras) determine the probability of tumor progression (Fearon and Vogelstein 1990; Kinzler and Vogelstein 1998). On the basis of the observation that mutations of the APC gene on chromosome 5q are found as frequently in small adenomatous polyps as in cancers, the loss of normalAPC function is considered an early (and possibly initiating) event in colorectal tumorigenesis (Powell et al. 1992; Morin et al. 1997). Products of the APC gene influence cell proliferation, adhesion, migration, and apoptosis (Kinzler and Vogelstein 1998). Activating mutations in codons 12 and 13 of the ras oncogene are important in the progression of adenomas but are not directly involved in malignant transformations in the bowel (Bos 1989; Ohnishi et al. 1997; Kinzler and Vogelstein 1998). Approximately 85 percent of colorectal cancers show inactivating mutations of the p53 tumor suppressor gene on chromosome 17p, resulting in loss of growth arrest and/or apoptosis; these mutations are important at a late stage in malignant transformation (Hollstein et al. 1991; Kinzler and Vogelstein 1998). Clonal expansion of colorectal tumors containing mutant p53 genes gains a selective survival advantage and becomes increasingly invasive and metastatic (Kinzler and Vogelstein 1998).

Because observational studies consistently show an association between cigarette smoking and adenomatous polyps (IARC 1986; Kikendall et al. 1989; Cope et al. 1991; Monnet et al. 1991; Zahm et al. 1991; Lee et al. 1993; Olsen and Kronborg 1993; Giovannucci et al. 1994b; Peipins and Sandler 1994; Boutron et al. 1995; Martínez et al. 1995; Longnecker et al. 1996; Nagata et al. 1999; Potter et al. 1999; Almendingen et al. 2000; Breuer-Katschinski et al. 2000; Inoue et al. 2000), Giovannucci and others have proposed that cigarette smoking plays a role early in colon and rectum carcinogenesis, likely acting on APC genes (Giovannucci et al. 1994a,b; Giovannucci and Martínez 1996). Two large cohort studies found that smoking for two decades or more was associated with large adenomas and that smoking for less than 20 years was associated with small adenomas (Giovannucci et al. 1994a,b). Cigarette smoking for at least three decades also has been associated with an increased risk of colorectal cancer incidence and mortality (Giovannucci et al. 1994a,b; Heineman et al. 1995; Chao et al. 2000). An initiating role of tobacco in the formation of adenomas is further supported by the finding that smokers who quit continue to have an elevated risk of adenoma recurrence after 10 years of smoking cessation (Jacobson et al. 1994). Cigarette smoking has not yet been associated with specific gene mutations or epigenetic changes associated with colorectal cancer.           

Cigarette smoke contains many carcinogens, including PAHs, heterocyclic aromatic amines, and N-nitrosamines (Hoffmann and Hoffmann 1997), that can reach the large bowel via the circulatory system or by direct ingestion of foods that contain these carcinogens (Giovannucci and Martínez 1996). One small study has documented that DNA adducts to metabolites of benzo[a]pyrene, a potent PAH, in colonic mucosa occur more frequently and at higher concentrations in smokers than in nonsmokers (Alexandrov et al. 1996). This study provides direct evidence that tobacco carcinogens bind to DNA in the human colonic epithelium. DNA adduction levels in the colonic epithelium have been found at higher levels in tumor tissue from colorectal cancer cases than from controls (Pfohl-Leszkowicz et al. 1995).

Other genes known to be important in colorectal cancer include mismatch repair genes associated with the hereditary familial syndrome, nonpolyposis colorectal cancer, and with sporadic cases of colorectal cancer (Liu et al. 1995, 1996; Thibodeau et al. 1998). One study has found that cigarette smoking is associated with a mismatch repair deficiency in colorectal cancers, reflected by a sixfold increased risk of microsatellite instability (a genetic marker) in tumors in current smokers compared with nonsmokers (Yang et al. 2000).

To date, the association between cigarette smoking and colorectal cancer has not been found to be modified by polymorphisms of genes important in the detoxification of carcinogens found in tobacco smoke, including glutathione S-transferase (GST) M1, T1, and N-acetyltransferase 2 (NAT2 ) (Gertig et al. 1998; Slattery et al. 1998). Studies of colorectal adenomas also have found no modification of the risk of cigarette smoking by polymorphisms of GSTM1, NAT2, or cytochrome P-4501A1, an enzyme important in the activation of PAHs (Lin et al. 1995; Potter et al. 1999; Inoue et al. 2000). However, one study found that when researchers examined only adenomas 1 cm or larger, current smokers with the GSTM1 null genotype were at a higher risk compared with those without the null genotype (Lin et al. 1995).

Animal Models

Animal models of tobacco carcinogenicity in the colon and rectum are limited and do not include studies in which the route of exposure is by inhalation. Adenocarcinomas of the colon have been produced in inbred male Syrian hamsters by intrarectal instillation of benzo[a]pyrene (Wang et al. 1985). In vivo mutational assay studies show that oral administration of benzo[a]pyrene to the lacZ transgenic mouse (Muta™ Mouse) induced the highest mutant frequency in the colon compared with other organs tested (Hakura et al. 1998, 1999; Kosinska et al. 1999). In vitro studies show that both rat and human colonic epithelium in cell cultures can enzymatically activate benzo[a]pyrene (Autrup et al. 1978).

Epidemiologic Evidence

Published studies on cigarette smoking and colorectal adenomatous polyps and cancer cited in this section were identified by searching the MEDLINE database from 1966 through July 2000 using the headings “tobacco,” “smoking,” “colorectal adenomas,” “colorectal neoplasms,” “colonic neoplasms,” and “rectal neoplasms,” and from the reference lists of published original and review articles in English on cigarette smoking and colorectal adenomas and cancer. The association between cigarette smoking and colorectal adenomas and cancer has been evaluated in a number of prospective and case-control studies since the 1960s. This review focuses on published studies that exclude cigar and pipe smokers, specify lifetime nonsmokers, and distinguish current from former smokers. If there are multiple reports from the same prospective cohort, results from the longest follow-up period are reported unless otherwise stated.

Table 2.26 presents prospective and retrospective studies of colorectal adenomatous polyps stratified by the cigarette smoking status of participants. Current cigarette smoking was consistently associated with an increased risk of colorectal adenomatous polyps in men and women, with OR estimates ranging between 1.5 and 3.8, adjusting for age and multiple covariates (Cope et al. 1991; Monnet et al. 1991; Zahm et al. 1991; Olsen and Kronborg 1993; Martínez et al. 1995; Longnecker et al. 1996; Nagata et al. 1999; Potter et al. 1999; Almendingen et al. 2000; Breuer-Katschinski et al. 2000; Inoue et al. 2000). Current smokers generally were at a higher risk compared with former smokers (Zahm et al. 1991; Martínez et al. 1995; Longnecker et al. 1996; Nagata et al. 1999; Potter et al. 1999; Almendingen et al. 2000; Breuer-Katschinski et al. 2000; Inoue et al. 2000). Former smokers had a significantly increased risk of colorectal adenomas compared with lifetime nonsmokers in five studies (Monnet et al. 1991; Olsen and Kronborg 1993; Martínez et al. 1995; Nagata et al. 1999; Potter et al. 1999), two of which also found an increased risk in former compared with current smokers (Monnet et al. 1991; Olsen and Kronborg 1993). One Japanese study found no increased risk of adenomas associated with current or former smoking (Kato et al. 1990b), and a randomized clinical trial of antioxidant vitamins in polyp prevention found no association between smoking and the recurrence of colorectal adenomas (Baron et al. 1998). Of two studies that compared adenoma cases to both hospital and population controls, one (Breuer-Katschinski et al. 2000) found an increased risk among current and former smokers only when comparing cases to hospital controls, whereas the other (Almendingen et al. 2000) found a comparably increased risk of adenomas among current and former smokers when comparing cases to either hospital or population controls.

Table 2.26

Epidemiologic studies on the association between smoking status and the risk of colorectal adenoma. 

Most studies examining the risk of adenomas in relation to cigarette smoking duration or pack-years have found a significantly positive association (Kikendall et al. 1989; Monnet et al. 1991; Zahm et al. 1991; Olsen and Kronborg 1993; Giovannucci et al. 1994a,b; Boutron et al. 1995; Martínez et al. 1995; Longnecker et al. 1996; Nagata et al. 1999; Potter et al. 1999; Almendingen et al. 2000; Inoue et al. 2000). Three prospective studies of the risk of proximal and distal colorectal adenomas have shown a significant dose-response relationship with total duration and with pack-years of smoking in men and women (Giovannucci 1994a,b; Nagata et al. 1999). Both the Health Professionals Follow-Up Study (Giovannucci et al. 1994b) and the Nurses Health Study (Giovannucci et al. 1994a) found that (1) smoking at least 20 years in the past was associated with the prevalence of large distal adenomas and (2) smoking fewer than 20 years was associated with small distal adenomas. Several case-control studies have reported a significant dose-response relationship with pack-years (Kikendall et al. 1989; Martínez et al. 1995; Longnecker et al. 1996; Potter et al. 1999) or with smoking duration (Olsen and Kronborg 1993; Almendingen et al. 2000) in studies of men and women combined. When examined separately by gender, there is a consistently significant dose-response relationship with pack-years and smoking duration among men (Monnet et al. 1991; Zahm et al. 1991; Lee et al. 1993; Boutron et al. 1995; Inoue et al. 2000) but a nonsignificant trend among women (Lee et al. 1993; Boutron et al. 1995). One case-control study reported no association between adenoma risk and pack-years in men or women (Sandler et al. 1993b).

Table 2.27 shows that cohort studies of colon and rectal cancer incidence and mortality among men in the United States consistently report an increased risk associated with current smoking status, with RRs ranging between 1.2 and 1.4 for colon cancer and between 1.4 and 2.0 for rectal cancer, regardless of the number or type of covariates adjusted for (Heineman et al. 1995; Chyou et al. 1996; Hsing et al. 1998; Chao et al. 2000; Stürmer et al. 2000). Two Norwegian studies also report risk estimates within this range (Tverdal et al. 1993; Engeland et al. 1996), but a study of Swedish male construction workers found no increased risk of colon cancer with current smoking (RR = 0.98) or former smoking (RR = 1.02) (Nyrén et al. 1996). More than half of the Swedish cohort was younger than 40 years of age at cohort entry, substantially younger than other cohorts in which an increased risk was observed. The 40-year follow-up of the British Physicians Study reported a RR of 1.36 for colon cancer mortality and 2.30 for rectal cancer mortality (Doll et al. 1994).

Table 2.27

Cohort studies on the association between current smoking and the risk of colorectal cancer incidence or mortality. 

CPS-II is the largest cohort study reporting an increased risk of colorectal cancer mortality associated with current smoking status in men (RR = 1.3) and women (RR = 1.4) (Chao et al. 2000). Two Norwegian cohort studies of women have found no increased risk associated with current smoking status (Tverdal et al. 1993; Engeland et al. 1996), similar to the eight-year follow-up report of the Nurses Health Study (Chute et al. 1991); two of these studies included women aged 30 through 55 years at enrollment (Chute et al. 1991; Tverdal et al. 1993). Two other cohort studies of men and women combined found no increased risk of colon or rectal cancer with cigarette smoking (Klatsky et al. 1988; Knekt et al. 1998). The RR estimates associated with former smoking among men and women fall within the range of 1.0 and 1.5 and, with some exceptions (Chute et al. 1991; Heineman et al. 1995; Engeland et al. 1996; Nyrén et al. 1996; Hsing et al. 1998), generally are intermediate between the risks observed among current smokers and lifetime nonsmokers.

Case-control studies of colon and rectal cancer incidence by cigarette smoking status generally have not reported an increased risk among male smokers (Table 2.28) (Kune et al. 1992; D’Avanzo et al. 1995; Le Marchand et al. 1997). The case-control studies are inconsistent for women alone and for women and men combined (Kune et al. 1992; Baron et al. 1994; D’Avanzo et al. 1995; Newcomb et al. 1995; Le Marchand et al. 1997). One study of U.S. women found significantly higher RRs in current smokers compared with lifetime nonsmokers, 1.3 for colon cancer and 1.7 for rectal cancer (Newcomb et al. 1995). When examined by cigarette smoking duration, the risk increased with the number of years the participants had smoked. The risks associated with having smoked 31 to 40 years were 1.7 for colon cancer and 1.5 for rectal cancer (Newcomb et al. 1995); it was the only study to adjust the risk estimates for colorectal cancer screening. Another study has examined the relationship by right and left colon and found a significantly increased risk of cancer in the right colon among former female smokers (OR = 2.4) and a nonsignificantly increased risk of cancer in the left colon and rectum among former male smokers compared with nonsmokers (Le Marchand et al. 1997). This study also reported a significantly increased risk of colon and rectal cancers associated with increments in pack-years of smoking in the distant and recent past among both genders (Le Marchand et al. 1997).

Table 2.28

Case-control studies on the association between smoking status and the risk of colorectal cancer incidence. 

Only more recent epidemiologic studies (since 1994) have examined colorectal cancer incidence or mortality in relation to gradients of smoking duration and timing, beyond smoking status (Giovannucci et al. 1994a,b; Nyrén et al. 1996; Hsing et al. 1998; Chao et al. 2000). Four recent reports from cohort studies have described an increased risk of colorectal cancer incidence and mortality with increased smoking duration in both men and women (Table 2.29) (Giovannucci et al. 1994a,b; Hsing et al. 1998; Chao et al. 2000). The sole exception is the Swedish study of men in whom no increased risk was observed with an increase in smoking duration (Nyrén et al. 1996). The Health Professionals Follow-Up Study (Giovannucci 1994b) reported a significantly increased risk among men who had smoked at least 40 to 44 years (RR = 1.7); the 16-year follow-up of the Nurses Health Study (Giovannucci 1994a) reported an elevated risk in women who had smoked more than 10 cigarettes a day for 35 to 39 years (RR = 1.5); and another cohort of U.S. men (Hsing et al. 1998) found an increased risk after smoking 20 to 29 years (RR = 2.4).

Table 2.29

Cohort studies on the association between the duration of current smoking and the risk of colorectal cancer incidence or mortality. 

CPS-II found a statistically significant increase in risk of colorectal cancer mortality among male smokers of 30 to 39 years’ duration (multivariate RR = 1.3) and among female smokers of 20 to 29 years’ duration (multivariate RR = 1.3) (Chao et al. 2000). Controlling for multiple covariates decreased age-adjusted estimates in currently smoking men but had little net effect on age-adjusted estimates in currently smoking women. Results of cohort studies that assess cigarette smoking status only at cohort enrollment may underestimate the true risk among long-term continuing smokers, because some smokers will have quit smoking during the cohort follow-up period.

Two cohort studies of colorectal cancer mortality have found a consistently increasing risk associated with a younger age at smoking initiation (Table 2.30) (Heineman et al. 1995; Chao et al. 2000). The 26-year follow-up of the veterans cohort reported that initiating smoking before 15 years of age was associated with a RR of 1.4 for colon cancer and 1.5 for rectal cancer (Heineman et al. 1995). CPS-II found that currently smoking men and women who began smoking at 15 years of age or younger had an increased risk of death from colorectal cancer (multivariate RR = 1.4 in men and 1.7 in women) (Chao et al. 2000).

Table 2.30

Cohort studies on the association between the age at initiation of current smoking and the risk of colorectal cancer mortality.

Data from CPS-II show that former smokers experience lower colorectal cancer mortality rates compared with continuing smokers (Table 2.31) (Chao et al. 2000). Risk decreases with a younger age at and a greater number of years since smoking cessation; former smokers who quit 20 or more years before the study were not at an increased risk of death from colorectal cancer compared with nonsmokers. Controlling for multiple covariates reduced the age-adjusted risk estimates in former male smokers but increased the risk estimates in former female smokers. The Leisure World cohort also found that men who had quit smoking more than 20 years ago were at a lower risk of colorectal cancer incidence than those who had quit within the past 20 years (Wu et al. 1987). In the multisite case-control study conducted by Slattery and colleagues (1997), risk remained modestly elevated for those former smokers who had stopped for 15 years or more.

Table 2.31

Cohort studies on the association between the number of years since or age at smoking cessation and the risk of colorectal cancer incidence or mortality. 

Evidence Synthesis

There is now a strong understanding of the sequence of genetic changes that leads from a normal cell to polyp development and then on to malignancy. Evidence points to an effect of smoking on polyp formation and possibly on the development of malignancy. Recent findings of prospective cohort studies suggest that long-term cigarette smoking is associated with an increased risk of colorectal cancer incidence and mortality in both men and women; risk is highest in current cigarette smokers, intermediate in former smokers, and lowest in nonsmokers. In some studies, the risk of colorectal cancer incidence and mortality tends to increase with longer smoking duration and a younger age at smoking initiation, and decreases with a younger age at and a greater number of years since successful smoking cessation, although the effects of these two factors cannot be readily separated because of their inherent correlation.

The aggregate epidemiologic evidence supports the hypothesis by Giovannucci and colleagues (1994a,b) and Giovannucci and Martínez (1996)that a latent period of several decades is necessary for cigarette smoking to increase colorectal cancer incidence or mortality, and that cigarette smoking likely plays a role in early colon and rectum carcinogenesis. This hypothesis is further supported by the association of smoking with adenomas. A number of studies show a greater risk for polyps in smokers compared with non-smokers, and some show a dose-response relationship with the number of cigarettes smoked. Under this hypothesis, the early studies of smoking might have missed an association because of insufficient follow-up time for the necessary tumor growth. This phenomenon would particularly apply to women, since the smoking epidemic began later in women than in men in the United States and most other developed countries. The finding of a declining risk following smoking cessation also suggests that cigarette smoking may affect later stages of the carcinogenic process leading to colorectal cancer.

In assessing whether cigarette smoking plays a causal role in colorectal cancer, consideration needs to be given to nutritional or other factors, such as physical activity and participation in colorectal cancer screening, that may confound the association. Not all recent studies have controlled for colorectal cancer risk factors that may be associated with smoking, such as physical inactivity. However, indirect evidence against confounding comes from the consistent finding of a small but statistically significant increase in risk associated with smoking, regardless of the set of covariates adjusted for in an analysis. Among the prospective cohort studies, three adjusted for physical activity or inactivity (Heineman et al. 1995; Chao et al. 2000; Stürmer et al. 2000). CPS-II analyses further adjusted for the use of estrogen replacement therapy (in women) and aspirin or other nonsteroidal anti-inflammatory drugs (Chao et al. 2000), factors that have been consistently associated with a lower risk of colorectal cancer (Thun et al. 1992; Calle et al. 1995; Potter 1999). Three cohort studies (Giovannucci et al. 1994b; Chao et al. 2000; Stürmer et al. 2000) adjusted for some measure of diet, and four studies (Giovannucci et al. 1994b; Hsing et al. 1998; Chao et al. 2000; Stürmer et al. 2000) adjusted for alcohol consumption. The only study of incidence or mortality that adjusted for screening sigmoidoscopy (as well as other variables) in women reported RR estimates similar to CPS-II results for smoking duration and years since quitting (Newcomb et al. 1995).

Adjusting for measured potential confounders for colorectal cancer in CPS-IIaffected the association with cigarette smoking differently by gender and by smoking status. Such adjustments increased risk estimates for former female smokers, had little net effect on risk estimates for current female smokers, and decreased the risk estimates for men. The slight decrease in adjusted estimates among men was comparable to that reported from the Health Professionals Follow-Up Study (Giovannucci 1994b), which controlled for saturated fat, folate, and dietary fiber and was one of the few studies that reported age- and multivariate-adjusted risk estimates. Although the possibility of residual confounding cannot be completely excluded, the internal consistency of findings and the fact that adjusting for measured potential confounders actually strengthened the association between smoking and colorectal cancer mortality in former female smokers in CPS-II suggest that the observed associations are unlikely to be explained solely by confounding. While the cohort study data are generally consistent with the hypothesis that smoking causes colorectal cancer, the trends of colorectal cancer incidence in the United States appear to be inconsistent. If smoking causes colorectal cancer after a substantial latent period as hypothesized (Giovannucci 2001), then the temporal patterns of smoking across the twentieth century would predict a decline in incidence in men before a decline in women. The opposite pattern has been observed (Ries et al. 2000b). However, other factors such as changes in risk variables and screening practices would also affect trends in incidence rates. Given the relatively modest effect of smoking on colorectal cancer risks, trends in incidence are an insensitive indicator of any trends in the effects of smoking over time.

Cigarette smoking is associated with a diagnosis of colorectal cancer at a more advanced stage of the disease (Longnecker et al. 1989), leading to a poorer prognosis and a lower survival rate in smokers compared with nonsmokers. However, recent cohort studies have reported similar findings of increased risks among smokers for both colorectal cancer incidence and mortality (Giovannucci et al. 1994a,b; Chao et al. 2000). Although no published reports were found on colorectal cancer screening prevalence by cigarette smoking status, the 1990–1994 National Health Interview Surveys (Rakowski et al. 1999) show that compared with lifetime nonsmokers, women who currently smoke are less likely, and those who are former smokers are more likely, to be screened for breast and cervical cancers. Thus, colorectal cancer mortality studies cannot exclude the possibility that continuing smokers experienced higher death rates from colorectal cancer than did nonsmokers because of less screening and a later stage of disease at diagnosis. However, the statistically significant increase in risk of colorectal cancer mortality among former female smokers in CPS-II argues against appreciable confounding by differential colorectal cancer screening practices, because these women are perhaps the most likely to be screened. CPS-II results were also similar to those of the one study that adjusted for screening sigmoidoscopy (Newcomb et al. 1995). The consistently observed relationship between cigarette smoking and adenomatous polyps, especially large adenomas (Kikendall et al. 1989; Cope et al. 1991; Monnet et al. 1991; Zahm et al. 1991; Lee et al. 1993; Olsen and Kronborg 1993; Giovannucci et al. 1994a,b; Peipins and Sandler 1994; Boutron et al. 1995; Martínez et al. 1995; Longnecker et al. 1996; Nagata et al. 1999; Potter et al. 1999; Almendingen et al. 2000; Breuer-Katschinski et al. 2000; Inoue et al. 2000), also suggests that confounding by screening is unlikely to explain the increased risk observed in studies of colorectal cancer incidence and mortality.

In 2000, about 23 percent of adults in the United States were current cigarette smokers, and 22 percent were former smokers (CDC 2002b). In 2001, 29 percent of high school students were current cigarette smokers (CDC 2002a). If long-term cigarette smoking is a cause of colorectal cancer (one of the most common cancers in western populations), the multivariate-adjusted RRestimates in CPS-II would indicate that about 12 percent of colorectal cancer deaths among men and 12 percent among women in the general population were attributable to smoking.

Cumulative findings from several recent, large prospective studies show an increased risk of colon and rectal cancer after smoking for two or more decades. The temporal pattern of the effects of smoking suggests that it may act in both earlier and later stages of carcinogenesis.

Conclusion

1. The evidence is suggestive but not sufficient to infer a causal relationship between smoking and colorectal adenomatous polyps and colorectal cancer.

Implications

The aggregate evidence suggests that cigarette smoking may be one of the avoidable factors that causes colorectal cancer. Current and former smoking should be included with other potential risk factors for this disease in clinical and public health settings, and further research should be directed at smoking and colorectal cancer risk.

The possible inclusion of colorectal cancer among the smoking-related cancers would substantially increase estimates of smoking attributable cancers and deaths worldwide. In the United States, the proportion of colorectal cancer deaths in 1997 attributable to any cigarette smoking (based on CPS-IImultivariate-adjusted RRs) would be approximately 12.0 percent among men and 12.3 percent among women, corresponding to an estimated 6,800 deaths. Considering past and future trends in cigarette smoking prevalence in the United States (Pierce et al. 1989) and in colorectal cancer incidence and mortality by gender since the 1950s (Chu et al. 1994), further reductions in smoking among adolescents and adults could accelerate and sustain future reductions in incidence and mortality.

Prostate Cancer

Prostate cancer is a leading cause of morbidity and mortality among men in the United States. It is more common in African American men than in white men, and the highest recorded rates in the world are among black men in the United States. In 2003, an estimated 220,900 new cases of prostate cancer were diagnosed, and an estimated 28,900 deaths were expected to occur (ACS 2003). Prostate cancer is the leading cause of cancer incidence among men (ACS 2003).

The risk of prostate cancer increases with age. African American men are at an increased risk, whereas Asian men are at a lower risk than white men. Lower vitamin A consumption and higher animal fat intake may increase the risk (Gann et al. 1994; Le Marchand et al. 1994), while a higher intake of lycopene may decrease the risk (Giovannucci et al. 1995; Giovannucci 1999). Having a vasectomy may be associated with an increased risk of prostate cancer 20 or more years after the procedure (Ross and Schottenfeld 1996). Endocrine factors, including testosterone and insulin-like growth factors, have been implicated in the development of this malignancy (Ross and Schottenfeld 1996; Giovannucci et al. 1997; Chan et al. 1998). Variations in the length of the androgen receptor gene CAG repeat may explain part of the excess risk in African American men (Platz et al. 2000).

Conclusions of Previous Surgeon General’s Reports

Previous Surgeon General’s reports have not addressed the relationship between smoking and prostate cancer.

Biologic Basis

During the last several decades there has been an explosion of epidemiologic studies addressing potential risk factors for this common malignancy, including cigarette smoking. Pathogenic mechanisms that may underlie the relationship between smoking and prostate cancer remain unclear. Carcinogens from tobacco can enter and concentrate in prostate cells (Smith and Hagopian 1981). Compared with men who do not smoke, men who smoke cigarettes have higher circulating levels of hormones formed in the adrenal gland (dehydroepiandrosterone, dehydroepiandrosterone sulfate, cortisol, and androstenedione) as well as testosterone, dihydrotestosterone, and sex hormone-binding globulin (Dai et al. 1988; Khaw et al. 1988; Field et al. 1994). This finding supports a potential mechanism for smoking because prospective epidemiologic studies have shown that testosterone is directly related to prostate cancer incidence and mortality (Nomura et al. 1988; Hsing and Comstock 1993; Gann et al. 1996).

Epidemiologic Evidence

The epidemiologic evidence relating smoking to the risk of prostate cancer has been mixed. Studies addressing disease incidence (which include case-control studies and several cohort studies) show an inconsistent increase in risk (Mishina et al. 1985; Honda et al. 1988; Hayes et al. 1994; van der Gulden et al. 1994), or no association between cigarette smoking and prostate cancer (Weir and Dunn 1970; Ross et al. 1987; Fincham et al. 1990; Talamini et al. 1992). Studies of mortality, largely limited to prospective cohort studies, show an increase in risk directly related to the number of cigarettes smoked. Investigators using different approaches to data analysis have attempted to determine whether this finding reflects a delayed diagnosis and treatment of smokers compared with nonsmokers, residual confounding factors, or a direct effect of tobacco smoke. Two studies found that smokers are more likely than nonsmokers to have their cancers diagnosed at a more advanced stage or histologic grade (Hussain et al. 1992; Daniell 1995).

Hsing and colleagues (1991) analyzed data from the follow-up of nearly 250,000 U.S. veterans and observed increased mortality rates for those who were current smokers at baseline. During 26 years of follow-up, approximately 4,600 men died of prostate cancer. Current smokers had a RRof 1.18 (95 percent CI, 1.09–1.28) compared with men who had never smoked, and the risk increased with the number of cigarettes smoked. Men smoking 40 or more cigarettes per day had a RR of 1.51 (95 percent CI, 1.20–1.90) compared with those who had never smoked. In this cohort, risks were higher during the first eight and one-half years of follow-up than during the remainder of the follow-up period, suggesting that recent smoking influenced the risk of prostate cancer mortality.

In an analysis of data from a follow-up of 348,874 men screened for the Multiple Risk Factor Intervention Trial, Coughlin and colleagues (1996)observed similar results. Compared with those who had never smoked, current smokers had a RR of 1.31 (95 percent CI, 1.13–1.52) for prostate cancer mortality. The risk increased with the number of cigarettes smoked; men smoking more than 25 cigarettes per day had a RR of 1.45 (95 percent CI, 1.19–1.97) compared with those who had never smoked.

The Lutheran Brotherhood Cohort Study also provides data on the association between smoking and prostate cancer. Hsing and colleagues (1990b) followed 17,633 white males for 20 years and documented 149 fatal cases of prostate cancer. The RR of prostate cancer mortality was significantly elevated for current smokers. Compared with men who had never smoked, smokers had a RR of 1.8 (95 percent CI, 1.1–2.9). Data from CPS-II were based on 1,748 deaths during nine years of follow-up of 450,279 men (Rodriguez et al. 1997). Current cigarette smoking was related to prostate cancer mortality in this cohort (RR = 1.34 [95 percent CI, 1.16–1.56]), but trends in risk were not observed with the number of cigarettes smoked per day or with the duration of smoking. Among 43,432 men in a prepaid health plan in northern California, Hiatt and colleagues (1994) observed similar results based on 238 deaths from prostate cancer. Men who smoked one or more packs of cigarettes per day had an adjusted RR that was 1.9 (95 percent CI, 1.2–3.1) compared with those who had never smoked.

The Health Professionals Follow-Up Study examined both incidence and mortality in an analysis of the association between smoking and prostate cancer, offering the possibility of considering issues related to etiology, delay in diagnosis, and mortality (Giovannucci et al. 1999). Lifetime cumulative smoking was unrelated to total prostate cancer incidence. However, men who had quit in the past 10 years were at an increased risk of diagnosis with distant metastatic prostate cancer (RR = 1.56 [95 percent CI, 0.98–2.48]) and fatal prostate cancer (RR = 1.73 [95 percent CI, 1.00–3.01]). Men who currently smoked cigarettes had an elevated risk of prostate cancer mortality; however, this risk was not statistically significant (RR = 1.58 [95 percent CI, 0.81–3.10]). Examining pack-years of cigarettes smoked in the preceding 10 years revealed a significant dose-response relationship with metastatic and fatal prostate cancer (p trend = 0.02). Men who smoked 15 or more pack-years in the preceding 10 years were at a higher risk of distant metastatic prostate cancer (RR = 1.81 [95 percent CI, 1.05–3.11]), and fatal prostate cancer (RR = 2.06 [95 percent CI, 1.08–3.90]) compared with nonsmokers. Within 10 years after smoking cessation, the excess risk was eliminated. In this cohort, the investigators also examined the relationship between smoking and survival after diagnosis. Men who smoked cigarettes had a lower survival rate than nonsmokers.

Several cohort studies do not show a significant increase in risk among cigarette smokers (Table 2.32). The British physicians cohort study found no clear association between smoking and prostate cancer mortality in 1951, 1957, 1966, 1972, 1978, and 1990. The heaviest smokers (smoking ≥25 cigarettes per day) had a RR of 1.24 for fatal prostate cancer compared with men who had never smoked (Doll et al. 1994). A similar association was observed among men followed for 20 years in Sweden (Adami et al. 1996). Current smokers had a RR for prostate cancer mortality of 1.26 (95 percent CI, 1.06–1.50) compared with men who had never smoked. Other studies with a single assessment of smoking status and follow-up periods of up to several decades did not show a clear association between smoking and prostate cancer (Whittemore et al. 1985; Carstensen et al. 1987; Severson et al. 1989).

Table 2.32

Cohort studies on the association between smoking status and behavior and the risk of prostate cancer incidence or mortality. 

Other Data

Differential screening and delay in seeking medical care have been hypothesized as possible explanations for the increased risk of prostate cancer mortality among cigarette smokers. In the study by Giovannucci and colleagues (1999), however, screenings for the prostate-specific antigen (PSA) did not differ substantially between groups. Among men younger than 65 years of age, 53 percent of those who had never smoked, 53 percent of the smokers who had quit in the past 10 years, and 50 percent of the current smokers had had at least one PSA test by 1994. For men 65 years of age or older the screening rates were higher: 79 percent of men who had never smoked, 78 percent of those who had quit in the past 10 years, and 70 percent of current smokers.

Smoking may relate to prostate cancer mortality through its impact on tumor characteristics. Two studies have suggested that smokers are more likely to have stage D tumors and to have poorly differentiated tumors (Hussain et al. 1992; Daniell 1995).

Evidence Synthesis

The suggestion of elevated risks for mortality and not for incidence (measured either in case-control studies or in prospective cohort studies) supports an association between smoking and prostate cancer mortality. The association between smoking and prostate cancer mortality rates appears to be reduced within 10 years of smoking cessation. The basis for this association is unclear. It might reflect more advanced disease in smokers, but evidence is limited.

If smoking contributed to the etiology of prostate cancer, an association of smoking with incidence would be anticipated, along with an increase in disease-specific mortality, assuming that cancers in smokers and nonsmokers are similar in clinical features.

Conclusions

1. The evidence is suggestive of no causal relationship between smoking and risk for prostate cancer.

2. The evidence for mortality, although not consistent across all studies, suggests a higher mortality rate from prostate cancer in smokers than in non-smokers.

Implications

Smoking cessation may reduce prostate cancer mortality. Further research is needed to refine this temporal relationship and to quantify the benefits of smoking cessation after diagnosis with prostate cancer.

Acute Leukemia

In 2003, an estimated 21,900 deaths attributable to leukemia and an estimated 30,600 new cases, evenly divided between acute and chronic leukemia, were expected to occur, affecting 10 times more adults than children (ACS 2003). In adults, the most common types of leukemia are acute myeloid (approximately 10,500 cases were diagnosed in 2003) and chronic lymphocytic (approximately 7,300 cases were diagnosed in 2003). Rates of acute myeloid leukemia among adults are higher in males than in females. In children, the most common type of leukemia is acute lymphocytic, accounting for 2,200 cases in 2003 (ACS 2003).

Conclusions of Previous Surgeon General’s Reports

The 1990 Surgeon General’s report (USDHHS 1990) noted that smoking has been implicated in the etiology of leukemia but the evidence was not consistent, and a conclusion was not reached regarding a possible causal relationship. The Surgeon General’s report on women and smoking (USDHHS 2001) concluded that acute myeloid leukemia has been consistently associated with cigarette smoking.

Biologic Basis

Several known leukemogenic substances are contained in cigarette smoke, including benzene and polonium-210 and lead-210 (which emit ionizing radiation). Both benzene and ionizing radiation (NRC 1990) are known causes of human leukemia that are associated with myeloid forms of leukemia and have little, if any, effect on the incidence of chronic lymphocytic leukemia. Radiation also causes acute lymphocytic leukemia in children (NRC 1990). Benzene, classified as a human carcinogen by IARC (1986), induces leukemia both in humans through occupational exposures and in laboratory animal models of this disease. Cigarette smoke is a major source of benzene exposure in the United States, accounting for roughly half of the exposures (Wallace 1996). Among smokers, 90 percent of benzene exposures come from smoking (Wallace 1996).

Data from human and experimental animal studies support the relationship between smoking and leukemia. Known leukemogens have been identified in cigarette smoke, and specific chromosomal abnormalities have been reported among smokers with leukemia. Sandler and colleagues (1993a) reported a higher frequency of smoking in persons with acute myeloid leukemia with specific chromosomal abnormalities (?7 or 7q?, ?Y, +13) than in similar patients without these abnormalities. In acute lymphoblastic leukemia the changes found in chromosomes were t(9;22) and (q34;q11).

Epidemiologic Evidence

A possible association between smoking and risk for leukemia was proposed by Austin and Cole (1986), who recommended further analyses of existing data to clarify the relationship between the amount smoked and specific forms of leukemia. Since then, numerous such analyses and new studies have been reported. By 1993, Siegel had systematically reviewed the literature, which included 21 published studies (including several reports from the follow-up of the same population), and concluded, after applying Hill’s causal criteria, that smoking was a cause of leukemia (Siegel 1993). Also in 1993, Brownson and colleagues reported a meta-analysis of published studies. They noted a significant association between current or former smoking and leukemia in general, and a stronger association between smoking and myeloid leukemia than with other subtypes (Brownson et al. 1993). Additional studies with similar findings have been published subsequently.

Both case-control and prospective cohort studies support the relationship between cigarette smoking and acute leukemia risk (Tables 2.33 and 2.34). The case-control approach affords the opportunity to quickly develop a series of cases for investigation and to uniformly classify the cases as to the type of leukemia. The results of case-control studies may be subject to information bias, arising from differential reporting of exposure by cases and controls. The prospective cohort studies do not have this limitation, but those using cause-specific mortality as the outcome measure may be affected by misclassification. In spite of these methodologic limitations, the evidence indicates an increased risk for leukemias in smokers. When risk estimates were provided by type, they tended to be higher for acute myeloid leukemia, usually called acute granulocytic leukemia or acute nonlymphocytic leukemia. A recent, large case-control study that included 807 persons with acute leukemia and 1,593 age-and gender-matched controls showed that the risk was highest among current smokers, and it decreased with years since smoking cessation (Kane et al. 1999).

Table 2.33

Case-control studies on the association between smoking and the risk of leukemia. 

Table 2.34

Cohort studies on the association between smoking and the risk of leukemia. 

The association appears stronger among the prospective cohort studies, although not all have shown a positive relationship (Table 2.34). The 20-year follow-up of the British physicians cohort study did not find an association (Doll and Peto 1978); however, with the 40-year follow-up, Doll and colleagues (1994) reported a significant dose-response association among cigarette smokers for myeloid leukemias but not for nonmyeloid leukemias. Men smoking 25 or more cigarettes per day had more than twice the age-standardized mortality rates of those who had never smoked.

In CPS-I, women who smoked had a lower risk of death from leukemia during the follow-up period than those who did not smoke (RR = 0.77) (Garfinkel and Boffetta 1990). A similar gender variation was reported by Friedman (1993) in the follow-up of participants enrolled in the Kaiser Permanente Medical Center multiphasic health check-up study. Among men, the RR of leukemia for current smokers was 2.8 (95 percent CI, 1.2–6.4); the RR for former female smokers compared with women who had never smoked was 0.9 (95 percent CI, 0.4–1.7). By contrast, CPS-II documented a significant positive association between former smoking and leukemia risks in women (RR = 1.34, p <0.05), and a significant dose-response relationship with the amount smoked in both women and men (Garfinkel and Boffetta 1990). These results were based on 327 deaths attributable to leukemia among men and 235 deaths among women.

McLaughlin and colleagues (1989) evaluated smoking and the 26-year risk of mortality from leukemia (based on 1,258 leukemia deaths) among the cohort of U.S. military service veterans for whom there were numerous follow-up reports (Hammond 1966; Kahn 1966; Rogot and Murray 1980; Kinlen and Rogot 1988). In the 26-year follow-up data, these authors found a significant relationship between smoking and all leukemias (with a dose-response association between the number of cigarettes smoked per day and the risk of leukemia). The strongest relationship was for myeloid leukemia (365 cases). The RR for current smokers of more than 20 cigarettes per day compared with persons who had never smoked was 1.95 (p <0.01). In this cohort study, which did not update smoking status after the baseline assessment, risk was stronger for the first 16 years of follow-up (RR = 1.6 [95 percent CI, 1.3–1.9]) than in the later 10 years (years 15 to 26 of the follow-up) (RR = 1.1 [95 percent CI, 0.9–1.3]) (McLaughlin et al. 1995a). In these data, the overall risk increased with the number of cigarettes smoked per day.

Cohort studies by Linet and colleagues (1991) and by Mills and colleagues (1990) also found a positive dose-response relationship between the number of cigarettes smoked and risk of leukemia. In the Lutheran Brotherhood Cohort Study, Linet and colleagues (1991) reported 74 deaths from leukemia (30 myeloid, 30 lymphatic, and 14 unspecified leukemia cases) among 17,633 white males followed for 20 years. The risk of total leukemia increased with the number of cigarettes smoked per day. Mills and colleagues (1990)followed 34,000 Seventh-Day Adventists for six years and identified 46 histologically-confirmed cases of leukemia. The group that had smoked the highest number of cigarettes in their lifetime had the highest risk of leukemia. These two cohorts were considerably smaller than the U.S. veterans and ACSstudies. Other studies supporting a positive dose-response relationship include some of the case-control studies.

Among the prospective studies, the 20-year follow-up of a cohort of construction workers in Sweden shows no relationship between smoking and leukemia (Adami et al. 1998). In this study, 400 cases of leukemia (including 171 myeloid leukemias) were diagnosed during follow-up. Current smokers had a RR for total leukemia of 1.0 (95 percent CI, 0.8–1.2) compared with workers who had never smoked. Similar null results were also observed for myeloid leukemia (RR = 1.0 [95 percent CI, 0.7–1.4]), and there was no evidence of a trend in risks with the number of cigarettes smoked per day.

Evidence Synthesis

A relationship between former or current smoking and the risk of acute myeloid leukemia is supported by evidence of a consistent dose-response relationship with the number of cigarettes smoked per day. The association of the duration of smoking with the degree of risk and an increase in risk among former smokers suggests that the relationship is not dependent on current smoking, but perhaps on the cumulative effects of cigarette smoking. This relationship is observed across diverse populations. The RR for persons who had ever smoked compared with non-smokers ranged from 1.3 to 1.5. Among those who smoked more than a pack of cigarettes per day the risk increased twofold. In 2002, IARC concluded that there is now sufficient evidence for a causal association between cigarette smoking and myeloid leukemia (IARC 2002).

Data from human and experimental animal studies provide evidence of a relationship between smoking and leukemia. Known leukemogens have been identified in cigarette smoke, and specific genetic alterations have been reported in smokers with leukemia. Benzene, a known leukemogen (Heath 1990), is found in cigarettes, and is the strongest known chemical leukemogen (Linet and Cartwright 1996). Polonium-210 and lead-210, alpha particle emitters in cigarette smoke, can reach the bone marrow where stem cells are located (Austin and Cole 1986; NRC 1988).

Korte and colleagues (2000) used risk assessment techniques for low-dose extrapolation to assess the proportion of leukemia and acute myeloid leukemia cases that could be attributed to the benzene in cigarettes. On the basis of linear potency models, these authors concluded that benzene in cigarette smoke contributed between 8 and 48 percent of smoking-induced leukemia deaths in total, and from 12 to 58 percent of smoking-induced acute myeloid leukemia deaths.

Conclusions

1. The evidence is sufficient to infer a causal relationship between smoking and acute myeloid leukemia.

2. The risk for acute myeloid leukemia increases with the number of cigarettes smoked and with duration of smoking.

Implications

The incidence of leukemia may remain elevated even after smoking cessation. Evidence is limited on the temporal pattern of change in risk after cessation, but a rapid decline in incidence has not been observed. Further research is needed to refine the patterns of risk after smoking cessation.

Liver Cancer

There are strong geographic variations in liver cancer incidence around the world. Although liver cancer is a relatively infrequent cause of cancer mortality in the United States, it is a leading cause of cancer deaths in the world (London and McGlynn 1996). In the United States, less than 1.5 percent of incident cancers are primary cancers of the liver and bile ducts. However, cancer of the liver ranks eighth (by deaths) on a worldwide basis, with three-quarters of the cases occurring in developing countries where hepatitis B and aflatoxin ingestion are prevalent causal exposures (Parkin et al. 1993). In the United States, an estimated 17,300 new cases of liver cancer and 14,400 deaths attributed to this cancer were expected to occur in 2003 (ACS 2003). Liver cancer is more common among men than women, in part reflecting the greater alcohol intake by men. Liver cancer incidence and mortality rates have increased since the 1980s in the United States (McKean-Cowdin et al. 2000). Hypotheses for this increase include the increasing frequency of hepatitis C virus and hepatitis B virus (HBV) infections.

Interpretation of the relationship between smoking and liver cancer is complicated by the potential for confounding by alcohol and HBV infections. First, alcohol intake is an established risk factor and smokers tend to drink more than nonsmokers, and this exposure has not been measured routinely in all studies that include information on smoking history. Second, chronic HBV infections are recognized as a major cause of this malignancy (IARC 1988). As for alcohol, not all epidemiologic studies that have addressed smoking have also assessed the hepatitis status of study participants. Hence, the unconfounded contribution of smoking to risks for liver cancer has been difficult to assess. Considerable epidemiologic evidence indicates, however, that smokers are at an increased risk for this cancer.

Conclusions of Previous Surgeon General’s Reports

The 1990 Surgeon General’s report (USDHHS 1990) noted an association between smoking and hepatocellular cancer that persisted after controlling for potentially confounding lifestyle factors including alcohol intake. That report also noted that HBV infections may modify the effects of smoking on the risk of liver cancer. The Surgeon General’s report on women and smoking (USDHHS 2001) concluded that smoking might be a contributing factor to the development of liver cancer.

Biologic Basis

Circulating carcinogens from tobacco smoke are metabolized in the liver, thus exposing the liver to many absorbed carcinogens. A long-term exposure to these carcinogens may therefore lead to cellular damage in the liver and the development of cancer. Carcinogens may act directly on the genes of the hepatocytes.

Epidemiologic Evidence

Epidemiologic data come from a wide range of studies in both low- and high-incidence countries (Table 2.35). Many of these studies have evaluated smoking, alcohol, and viral causes of liver cancer thoroughly, although some of the larger cohort studies have not controlled for each of these causal agents in assessing smoking’s effect. Cigarette smoking was directly related to the risk of liver cancer as the number of cigarettes smoked per day increased in some case-control studies (Yu et al. 1983; Trichopoulos et al. 1987b; Kuper et al. 2000) but not in others (Tanaka et al. 1992). In a cohort study of U.S. veterans, Hsing and colleagues (1990a) noted a significant trend in increased risks with an increasing number of cigarettes smoked, but their analysis did not control for alcohol consumption or hepatitis viral status. On the other hand, Doll and colleagues (1994) did not observe a trend in risk with higher levels of cigarette smoking in the 40-year report of the British physicians cohort study, and concluded that smoking is not related to liver cancer. In a 12-year cohort study of 14,397 residents of Taiwan aged 40 years and older, cigarette smoking was positively related to mortality from liver cancer (Liaw and Chen 1998). Among men, 110 deaths from liver cancer were identified, and for current smokers the RR was 2.2 (95 percent CI, 1.4–3.6) compared with persons who had never smoked. These authors adjusted for alcohol consumption and the presence of HBV surface antigens.

Table 2.35

Studies on the association between smoking and the risk of liver cancer. 

For persons smoking more than a pack a day, the RR for liver cancer has been 2 or more in both case-control and cohort studies, compared with the risk for persons who had never smoked (Yu et al. 1983; Hsing et al. 1990a; Doll et al. 1994; Kuper et al. 2000). However, not all studies have found an effect of this magnitude (Tanaka et al. 1992; Chiesa et al. 2000; Mori et al. 2000a). This inconsistency may be in part due to the study design and to the relative contribution of HBV infection to the risk of malignancy. For example, Lam and colleagues (1982) observed a RR of 3.3 (95 percent CI, 1.0–13.4) among current smokers, but the association was confined to those who were HBV-negative. Similarly, Trichopoulos and colleagues (1980, 1987b) observed significant associations among HBV-negative persons. In contrast, in a cohort of HBV-positive men and women in China, Tu and colleagues (1985)observed a RR of 4.6. One explanation for the varying results is the dominant role of hepatitis viral infection and the extent to which its effects have been considered in the studies on smoking. The higher RRs that were observed in several studies of persons who were negative for HBV compared with those who were positive suggest that this explanation is plausible.

Evidence Synthesis

A substantial body of epidemiologic evidence supports a relationship between smoking and liver cancer, but a positive association was not found in all studies considered. The metabolism in the liver of the many carcinogens from tobacco smoke leads to an exposure of hepatocytes to these carcinogens. The strength of an association between cigarette smoking and liver cancer varies according to HBV infection status, with stronger associations among those who are negative for HBV. In many of the studies, risk increases with the number of cigarettes smoked per day. Although confounding by alcohol and HBV infection status may bias the findings of some studies, controlling for these causes does not remove the strong association between smoking and liver cancer seen in several of the studies summarized in this report. Finally, in 2002, IARC concluded that there is now sufficient evidence for a causal association between cigarette smoking and cancer of the liver (IARC 2002).

Conclusion

1. The evidence is suggestive but not sufficient to infer a causal relationship between smoking and liver cancer.

Implications

The global burden of liver cancer may increase if smoking increases around the world. Further research is needed to resolve the relationship of smoking to liver cancer with further consideration of the history of hepatitis infection and alcohol use.

Adult Brain Cancer

Brain cancer incidence is higher in men than in women. In 2003, an estimated 18,300 new cases (10,200 among men and 8,100 among women), and an estimated 13,100 deaths attributed to brain cancer were expected to occur (ACS 2003).

The systematic epidemiologic study of brain cancer is hampered by the grouping of clinicopathologic entities and by problems with the accurate diagnosis of intracranial lesions. Further, it often is difficult to distinguish primary from secondary or metastatic lesions. Risk factors for brain cancers include working in petrochemical, rubber, and agricultural industries. Radiation exposure also has been related to the risk of brain cancer (NRC 1990; Preston-Martin and Mack 1996).

Conclusions of Previous Surgeon General’s Reports

Previous Surgeon General’s reports have not reviewed brain cancer and smoking.

Biologic Basis

Exposure to nitroso compounds has been related to the risk of brain cancer, stimulating interest in cigarette smoke as a source of exposure. Two major subcategories of nitroso compounds include nitrosamines, which require metabolic activation, and nitrosamides, which do not. The nitrosamides, particularly nitrosoureas, are effective nervous system carcinogens in many species (Preston-Martin and Mack 1996). Nitrosamides have been shown to damage DNA by the production of adducts. The major sources of exposure to nitrosamines in the United States are tobacco smoke, cosmetics, automobile interiors, and cured meats.

Epidemiologic Evidence

Both case-control and cohort studies have evaluated the relationship between smoking and cancer of the brain. In the 26-year follow-up of the U.S. veterans cohort (Hsing et al. 1991), no relationship was observed between smoking and mortality from brain cancer. In a population-based case-control study in Los Angeles County, California, that included 94 women with intracranial gliomas, no relationship was observed between cigarette smoking and the risk of brain cancer (Blowers et al. 1997). In a comparable study from the San Francisco Bay area that included 434 adults with incident glioma, men but not women were at an increased risk of cancer if they had smoked unfiltered cigarettes. Among the men, those who reported using filter-tipped cigarettes had no increase in risks compared with men who had never smoked (RR = 0.8 [95 percent CI, 0.5–1.2]), and those who smoked unfiltered cigarettes had an increased RR of 1.8 (95 percent CI, 0.9–3.4) (Lee et al. 1997). Among the women, an increased risk was not observed, although the prevalence of smoking unfiltered cigarettes was substantially lower. An Australian case-control study also failed to show any relationship between smoking and glioma in women, but did show a suggestive relationship in men (Ryan et al. 1992). On the basis of 416 cases (166 women and 250 men), Hurley and colleagues (1996) reported that men who had smoked had a RR for glioma of 1.64 (95 percent CI, 1.10–2.45) compared with men who had never smoked, while for women who had smoked the RR was 0.99 (95 percent CI, 0.62–1.62) compared with women who had never smoked. In this study, there was no evidence of an increase in risk among either women or men with increased durations of smoking or pack-years of smoking.

Eight other studies, all smaller than those reviewed above, have also failed to find an association between smoking and glioma (Musicco et al. 1982; Ahlbom et al. 1986; Burch et al. 1987; Brownson et al. 1990; Hochberg et al. 1990; El-Zein et al. 1999; Bondy et al. 2001; Zheng et al. 2001). In several of these studies, controls were limited to hospitalized patients— a potential source of bias when evaluating smoking-related risks (Musicco et al. 1982; Burch et al. 1987). Ahlbom and colleagues (1986) studied 78 cases and observed no association between smoking and astrocytoma when using population controls (RR = 1.2 [95 percent CI, 0.6–2.5]). Musicco and colleagues (1982) observed a nonsignificant increase in risk when comparing heavy smokers with persons who had never smoked (RR = 1.5, p = 0.71). Burch and colleagues (1987) compared 215 cases with 215 hospital controls, and observed an overall RR of 1.44 (95 percent CI, 0.94–2.21) comparing smokers of plain cigarettes with nonsmokers, and a RR of 0.98 (95 percent CI, 0.66–1.46) comparing smokers of filter-tipped cigarettes with nonsmokers. There was a significant increase in risk with an increased amount smoked for those smoking plain cigarettes (p = 0.026) but not for those smoking filter-tipped cigarettes (p = 0.64).

Evidence Synthesis

Overall, the epidemiologic evidence shows no consistent relationship between smoking and glioma. Duration of smoking, the number of cigarettes smoked per day, and pack-years of smoking have been evaluated in different studies. None of these measures of exposure shows a strong or consistent relationship.

Conclusion

1. The evidence is suggestive of no causal relationship between smoking cigarettes and brain cancer in men and women.

Implications

Epidemiologic research using both case-control and cohort designs has not found an association between smoking and brain cancer in adults. Any new studies on this topic will need to have large sample sizes and careful characterizations of the tumors.

Disclaimer: The author of each article published on this web site owns his or her own words. The opinions, beliefs and viewpoints expressed by the various authors and forum participants on this site do not necessarily reflect the opinions, beliefs and viewpoints of Utah Standard News or official policies of the USN and may actually reflect positions that USN actively opposes. No claim in public domain or fair use.    © Edmunds Tucker. UTopiAH are trade marks of the author. Utopia was written in 1515 by Sir Thomas More, Chancellor of England. A word of wisdom.

[ 2004 Health Consequences of Smoking: the Surgeon General continues next at Part  96.]