Cigarette smoking, use of other tobacco products and stomach cancer mortality in US adults: The Cancer Prevention Study II

Authors


Abstract

Cigarette smoking is associated with increased risk of stomach cancer in many studies but there are limited data on this relationship in women and on risk associated with use of tobacco products other than cigarettes. We examined stomach cancer death rates in relation to cigarette smoking in women and use of cigarette, cigar, pipe, or smokeless tobacco in men in a nationwide prospective mortality study in the United States (US). Cohort follow-up from 1982–96 identified 996 and 509 stomach cancer deaths among 467,788 men and 588,053 women, respectively. Cox proportional hazards models were fitted to estimate rate ratios (RR) and 95% confidence intervals (CI) using non-users of tobacco as the referent group. Multivariate-adjusted RRs were the highest for men who currently smoked cigars (RR = 2.29, 95% CI = 1.49–3.51) or cigarettes (RR = 2.16, 95% CI = 1.75–2.67) and both increased with smoking duration. Women who currently (RR = 1.49, 95% CI = 1.18–1.88) or formerly (RR = 1.36, 95% CI = 1.08–1.71) smoked cigarettes were at significantly increased risk, as were men who formerly smoked cigarettes (RR = 1.55, 95% CI = 1.28–1.88), or currently (RR = 1.81, 95% CI = 1.40–2.35) or formerly (RR: 1.57, 95% CI = 1.22–2.03) used more than one type of tobacco. Men who reported a history of chronic indigestion or gastroduodenal ulcer had substantially higher mortality rates associated with current cigarette (RR = 3.45, 95% CI = 2.05–5.80) or cigar (RR = 8.93, 95% CI = 4.02–19.90) smoking, as did men who were current aspirin users. If causal, the estimated proportion of stomach cancer deaths attributable to tobacco use would be 28% in US men and 14% in women. We conclude that prolonged use of tobacco products is associated with increased stomach cancer mortality in men and women. The accumulated evidence from this and other studies support reconsidering stomach cancer as a tobacco-related cancer. © 2002 Wiley-Liss, Inc.

Over 1 billion people use tobacco worldwide.1 Consumption of tobacco products is increasing in many parts of Asia, Latin America and Eastern Europe.2, 3 In the United States (US), cigar smoking and snuff dipping also have increased in recent years, especially among youth and young adults.4, 5 Tobacco is projected to be the single most important and preventable cause of disease in the world by the 2020s;6 it already is in the US and Western Europe.7 Eight cancers have been designated by the US Surgeon General to be caused by tobacco smoking.8 Stomach cancer was counted among the smoking-attributable cancers in 19879 but subsequently removed from the list due to uncertainties about causality;8 no consistent dose-response with smoking duration or intensity had been shown in studies published up to the late 1980s.10, 11

Stomach cancer is the second most common cancer globally.12, 13 The highest risk is seen in areas of the world where chronic infection with the bacterium Helicobacter (H.) pylori remains endemic.14, 15, 16 If tobacco use causes stomach cancer, then the emerging and maturing tobacco epidemic in intermediate- to high-risk areas for stomach cancer will likely contribute to substantial increases in stomach cancer incidence and mortality worldwide.

In the US, overall rates of stomach cancer have decreased during the past century, coincident with a decline in the prevalence of H. pylori infection and other important risk factors.15, 16 Since the 1970s, however, incidence of adenocarcinomas of the gastric cardia has increased rapidly.17, 18 These tumors now represent nearly half of all stomach cancers in US men16, 17 and appear not to be H. pylori-related.14, 15, 19, 20 Prolonged cigarette smoking has been postulated to contribute to the increase in gastric cardia adenocarcinomas in the US and Western Europe, reflecting a lag time of 30 or more years between tobacco use and clinical disease.21

Cigarette smoking has been positively associated with stomach cancer incidence and mortality in prospective22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32 and case-control studies33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56 of men, although not all studies have found such an association57, 58, 59, 60, 61, 62, 63, 64 and not all positive studies have observed a consistent dose-response relationship with smoking duration or intensity. Prospective studies of women in Scandinavia have not found a positive association28, 58, 59 and results from case-control studies of women have been inconsistent.37, 40, 47, 51, 52 Most published studies, especially prospective studies, have not adjusted for potentially important covariates such as socioeconomic status, dietary factors, or H. pylori infection.65, 66 Data on the use of cigar, pipe, or smokeless tobacco in relation to stomach cancer are limited. Only a few studies have stratified the analyses by H. pylori infection status.

We examined stomach cancer mortality in relation to the use of 4 types of tobacco products (cigarettes, cigars, pipe and smokeless tobacco including chewing tobacco and snuff) in men and cigarette smoking in women in the Cancer Prevention Study II (CPS II) prospective cohort, adjusting for multiple potential confounding factors.

MATERIAL AND METHODS

Study population

CPS II is a prospective mortality cohort established by the American Cancer Society (ACS) in 1982, with 508,351 men and 676,306 women recruited by ACS volunteers in 50 states, the District of Columbia and Puerto Rico. To be eligible for the study, individuals had to be age 30 years or older and reside in a household in which at least 1 person was age 45 years or older. Median age at cohort entry was 57 years for men and 56 years for women.

This analysis is based on 14 years of cohort follow-up from study enrollment in 1982 to December 31, 1996. Vital status of study participants was obtained by ACS volunteers in 1984, 1986 and 1988; reported deaths were verified with death certificates. Since 1988, vital status has been ascertained via automated linkage of the CPS II study population with the National Death Index (NDI) for date and, since 1993, cause of death.67 The underlying cause of death, from death certificate or NDI, was classified according to the International Classification of Disease (ICD), 9th Revision.68 Our outcome of interest was death due to stomach cancer (ICD-9 codes 151.0–151.9). As of December 31, 1996, 129,642 men and 107,810 women had died and 2,897 (0.24%) individuals had follow-up truncated in 1988 due to insufficient information for NDI linkage. We excluded from this analysis individuals who reported a previous cancer (other than non-melanoma skin) at cohort enrollment (25,242 men and 57,107 women) and those with missing or uninterpretable tobacco use information (15,321 men and 31,146 women). The final analytic cohort after exclusions consisted of 467,788 men and 588,053 women and 996 and 509 deaths from stomach cancer, respectively.

Tobacco use information

At enrollment in 1982, study participants completed a questionnaire that included questions on current and past use of multiple types of tobacco products in men (cigarettes, cigars, pipes, chewing tobacco and snuff) and cigarettes in women. CPS II men were grouped into never, current and former users of each type of tobacco at enrollment. Due to small numbers, men who exclusively chewed tobacco or used snuff were grouped together as smokeless tobacco users. Men who reported having used more than one type of tobacco were considered separately as current users if they reported using any tobacco product at enrollment. Lifelong non-users of any tobacco were the referent group for all analyses.

Smokers of exclusively cigarettes or cigars were further characterized by years of use, average number of cigarettes or cigars smoked per day, age started smoking, depth of cigar smoke inhalation (do not inhale to inhale deeply) and, among former smokers, age quit smoking and number of years since quit. Categories with <20 stomach cancer deaths (pipe and smokeless tobacco) were not examined beyond status of use. Current and former users of more than one type of tobacco were characterized by the number of years between first and last use of any tobacco, number of years since first use and, among former users, age quit using any tobacco and the number of years since quit. CPS II women were grouped into never, current and former cigarette smokers at enrollment, with smoking characteristics examined similarly as men.

Covariate information

Potential confounders were chosen based on their observed association with stomach cancer mortality69, 70 and with tobacco use in CPS II and on stomach cancer risk factors reported in the literature.71, 72 Covariates included in final multivariate models were age (single year), race (white, black, other, or unspecified), education (some high school, high school graduate, some college or trade school, college graduate or post-graduate work, or unspecified), family history of stomach cancer among first degree relatives (yes, no), intake of high-fiber grain foods (in quintiles) and vegetables (in quintiles), consumption of citrus fruits or juices (in tertiles), a variable combining supplemental vitamin C use frequency (never, occasional, 1–29 per month, 30+ per month, or unspecified) and duration (never, 1–3 years, 4–9 years, or 10+ years), a variable combining supplemental multivitamin use frequency (never, occasional, 1–29 per month, or 30+ per month) and duration (never, 1–4 years, 5–9 years, or 10+ years) and a variable combining aspirin use frequency (never, occasional, 1–14 per month, 15+ per month, or unspecified) and duration (never, 1–9 years, 10–19 years, 20+ years, or unspecified).

The diet section of the questionnaire included 32 foods and 10 beverages. Food intake variables were derived by summing the number of days per week that each participant reported consuming each vegetable (carrots, tomatoes, potatoes, squash/corn, green leafy vegetables, raw vegetables, cabbage/broccoli/Brussels sprouts), high-fiber grain foods (bran/corn muffins, brown rice/whole wheat/barley and oatmeal/shredded wheat/bran cereals) and citrus fruits or juices.

The following variables were also evaluated as potential confounders: body mass index, exercise, alcohol consumption, own or parents' country of birth (as a crude proxy for early childhood exposures such as socioeconomic status or H. pylori) and personal history of chronic indigestion or gastroduodenal ulcer. None of these materially altered risk estimates associated with tobacco use and thus were not included in final multivariate models. These variables were evaluated as potential effect modifiers, as were family history of stomach cancer and aspirin use.

Statistical methods

Age-adjusted stomach cancer mortality rates were directly standardized to the age distribution of the 1970 world population age 30 years and over. Maximum likelihood estimates of rate ratios (RR) and 95% confidence intervals (CI) were obtained by fitting Cox proportional hazards models73 using lifelong nonusers of tobacco as the referent group. p-Values for trend were estimated by modeling tobacco use characteristics as continuous variables, excluding lifelong nonusers.74 The Wald statistic was used to test for homogeneity of stratum-specific rate ratios75 in analyses stratified by covariates. The proportion of stomach cancer deaths attributable to tobacco use in the population was estimated based on CPS II multivariate-adjusted RRs associated with ever use of any tobacco product and prevalence estimates of any tobacco (cigarette, cigar, pipe, or smokeless tobacco) use in US adult men and cigarette smoking women in 1991.4 All analyses were carried out with the SAS software package;76 all statistical tests were 2-sided.

RESULTS

Patterns and correlates of tobacco use

Seventy-five percent of CPS II men reported ever using any tobacco product; 32% were current and 38% former users of at least one tobacco product at enrollment, 5% were ever users with incomplete information. Twenty percent of men were current and 27% former smokers of exclusively cigarettes, less than 2% currently and exclusively smoked cigars. Seventeen percent of men reported having used more than one type of tobacco in their lifetime; of these, over 61% reported smoking cigarettes in combination with cigars or pipe, 25% smoked cigarettes and used smokeless tobacco and 14% used smokeless tobacco and smoked cigars or pipe. Nearly 21% of women were current cigarette smokers and another 21% former smokers in 1982.

Current users of any tobacco were more likely to report low or no intake of high-fiber grain foods, vegetables and citrus fruits or juices compared to lifelong non-users (Table I). With the exception of pipe smokers, male current users of tobacco products were less likely to report education beyond high school compared to non-users. Men who currently used any tobacco were less likely to take supplemental vitamin C or multivitamins than never or former users. Women who quit smoking cigarettes reported higher educational attainment and higher intake of high-fiber grain foods, vegetables and citrus fruits or juices compared to never- or current-smoking women; they were also more likely to report current use of supplemental vitamin C, multivitamins, or aspirin. Women who were current cigarette smokers were less likely to report consumption of high-fiber grain foods, vegetables and citrus fruits or juices than never- or former-smokers.

Table I. Study Participant Characteristics by Selected Tobacco Use Status at Study Baseline1
Tobacco use statusTotal nRace: non-whiteHigh school graduate or lessFamily history of stomach cancerLow or no intake of vegetablesLow or no intake of citrus fruits or juicesNo intake of high-fiber grain foodsCurrent user of multi-vitaminsCurrent user of vitamin C supplementCurrent use of aspirin (≥15/month)
  • 1

    CPS II men and women, 1982. Data in %.

Men          
 Never user117,968630415241831257
 Current cigarette only93,468841322403027197
 Current cigar only6,945635421292725187
 Current pipe only8,460426316272129217
 Current chewing tobacco or snuff2,855666424392621167
 Current user, >1 type of tobacco39,678444418352427209
 Former cigarette only126,6044344152821302310
 Former user, >1 type of tobacco40,5903284132416342711
Women          
 Never smoker326,835746319281836268
 Current cigarette122,465845323382536258
 Former cigarette122,4555343142616423110

Stomach cancer mortality by type of tobacco use

Ever users of any tobacco product were at significantly higher risk of stomach cancer death compared to never users (multivariate-adjusted RR = 1.68, 95% CI = 1.42–1.98 for men and RR = 1.38, 95% CI = 1.15–1.65 for women); risk estimates for current use were consistently higher than those for former use (Table II). Men who currently smoked cigars (RR = 2.29, 95% CI = 1.49–3.51) or cigarettes (RR = 2.16, 95% CI = 1.75–2.67) were at the highest risk of stomach cancer death, followed by men who formerly smoked cigarettes (RR = 1.55, 95% CI = 1.28–1.88) and currently (RR = 1.81, 95% CI = 1.40–2.35) or formerly (RR = 1.57, 95% CI = 1.22–2.03) used more than one type of tobacco. Current pipe smokers, users of smokeless tobacco and former cigar smokers had smaller increases in risk. Women who currently (RR = 1.49, 95% CI = 1.18–1.88) or formerly (RR = 1.36, 95% CI = 1.08–1.71) smoked cigarettes had significantly higher stomach cancer mortality compared to women who never smoked. Adjustment for multiple covariates somewhat elevated RRs in former users and slightly reduced the RRs but did not eliminate the statistically significant increased risk in current users.

Table II. Stomach Cancer Mortality Rates1
Type of tobacco useNumber of study participantsPerson-yearsStomach cancer deathsMortality rate per 100,0002Age-adjusted RR (95% CI)Multivariate-adjusted RR (95% CI)3
  • 1

    Age- and multivariate-adjusted mortality rate ratios (RR) and 95% confidence intervals (CI) by tobacco use status at study baseline, CPS II men and women, 1982 through 1996.

  • 2

    Mortality rates directly standardized to the 1970 world population of men and women age 30 years and over.

  • 3

    Multivariate models include age, race, education, family history of stomach cancer, consumption of high-fiber grain foods, vegetables, citrus fruits or juices, and use of vitamin C, multivitamins, and aspirin.

  • 4

    Mortality rates not provided for categories with fewer than 20 stomach cancer deaths.

MEN467,7885,980,0519967.1  
 Never-user of any type of tobacco117,9681,558,5521695.21.00 (Referent)1.00 (Referent)
 Ever-user of any type of tobacco349,8204,421,4998277.61.73 (1.47–2.04)1.68 (1.42–1.98)
  Current user of any type of tobacco151,4061,903,5153758.72.12 (1.76–2.54)1.97 (1.64–2.37)
   Current-cigarette only93,4681,166,2802309.62.41 (1.96–2.96)2.16 (1.75–2.67)
   Current-cigar only6,94590,4352510.72.47 (1.62–3.78)2.29 (1.49–3.51)
   Current-pipe only48,460110,34116 1.30 (0.78–2.18)1.33 (0.80–2.23)
   Current-chewing tobacco/snuff only42,85535,6738 1.79 (0.88–3.65)1.58 (0.76–3.28)
   Current->1 type of tobacco39,678500,785968.31.85 (1.44–2.38)1.81 (1.40–2.35)
  Former user of any type of tobacco179,8332,298,1244006.91.49 (1.24–1.78)1.51 (1.26–1.81)
   Former-cigarette only126,6041,618,0132857.11.56 (1.29–1.89)1.55 (1.28–1.88)
   Former-cigar only45,83473,74413 1.25 (0.71–2.19)1.28 (0.72–2.25)
   Former-pipe only45,92478,0286 0.64 (0.29–1.45)0.70 (0.31–1.60)
   Former-chewing tobacco/snuff only488110,8122 1.21 (0.30–4.92)1.11 (0.27–4.50)
   Former->1 type of tobacco40,590517,527945.61.44 (1.12–1.85)1.57 (1.22–2.03)
  Ever-user, incomplete information18,581219,860529.81.65 (1.21–2.26)1.52 (1.10–2.10)
WOMEN588,0537,913,5435092.6  
 Never smoked cigarettes326,8354,404,9442822.31.00 (Referent)1.00 (Referent)
 Ever smoked cigarettes261,2183,508,5992273.01.34 (1.12–1.60)1.38 (1.15–1.65)
  Current-cigarette only122,4651,629,2961053.21.51 (1.20–1.90)1.49 (1.18–1.88)
  Former-cigarette only122,4551,665,4261063.01.26 (1.01–1.59)1.36 (1.08–1.71)
  Ever smoker, incomplete information416,298213,87716 1.16 (0.70–1.92)1.10 (0.67–1.83)

Stomach cancer mortality rates by birth year and tobacco use

Among men, age-specific death rates from stomach cancer were higher at all ages in current users of any tobacco at the time of enrollment than in lifelong non-users of tobacco and higher in men born in earlier calendar years (Fig. 1). The patterns among women were similar but less stable due to small numbers (data not shown).

Figure 1.

Age-specific stomach cancer death rates by tobacco use status at study enrollment and birth year, CPS II men, 1982 to 1996

Dose-response analysis

Among current cigarette smokers (Table III), the multivariate-adjusted RRs increased with smoking duration (men, p-trend = 0.059), average number of cigarettes smoked per day (women, p-trend = 0.038), younger age at smoking initiation (men, p-trend = 0.075) and with pack-years (women, p-trend = 0.053). In currently smoking men, the RR estimates increased with number of cigarettes smoked per day except in the highest category of 2 or more packs per day.

Table III. Dose-Response Gradients in Current Smokers of Exclusively Cigarettes1
 MenWomen
nAge-adjusted RR (95% CI)Multivariate-adjusted RR (95% CI)2nAge-adjusted RR (95% CI)Multivariate-adjusted RR (95% CI)2
  • 1

    Number of stomach cancer deaths (n), age- and multivariate-adjusted rate ratios (RR) and 95% confidence intervals (CI), CPS II men and women, 1982 through 1996.

  • 2

    Multivariate models include age, race, education, family history of stomach cancer, consumption of high-fiber grain foods, vegetables, citrus fruits or juices, and use of vitamin C, multivitamins, and aspirin.

  • 3

    Estimates of p trend excluded non-users of tobacco.

Never user1691.00 (Referent)1.00 (Referent)2821.00 (Referent)1.00 (Referent)
Cigarette smoker2302.41 (1.96–2.96)2.16 (1.75–2.67)1051.51 (1.20–1.90)1.49 (1.18–1.88)
Years smoked cigarettes
 <2041.34 (0.46–3.91)1.17 (0.40–3.37)71.53 (0.72–3.26)1.43 (0.67–3.04)
 20–2991.08 (0.53–2.20)0.97 (0.48–1.98)141.26 (0.72–2.22)1.23 (0.70–2.17)
 30–39672.23 (1.62–3.06)2.08 (1.51–2.87)391.83 (1.28–2.61)1.84 (1.29–2.64)
 ≥401282.58 (2.04–3.28)2.36 (1.84–3.02)351.43 (1.00–2.04)1.47 (1.02–2.10)
 p trend3 0.04820.0589 0.14930.0743
Cigarettes smoked per day
 <20512.00 (1.46–2.74)1.66 (1.21–2.29)391.35 (0.97–1.90)1.31 (0.93–1.84)
 20 (1 pack)812.83 (2.16–3.70)2.51 (1.90–3.31)281.26 (0.85–1.86)1.25 (0.84–1.85)
 21–39562.75 (2.02–3.75)2.73 (1.99–3.75)182.05 (1.26–3.32)2.15 (1.32–3.50)
 ≥40 (≥2 packs)351.89 (1.30–2.74)1.83 (1.25–2.67)122.11 (1.17–3.78)2.15 (1.20–3.87)
 p trend3 0.79040.5392 0.07850.0375
Age started smoking cigarettes, years
 ≤15562.50 (1.84–3.39)2.17 (1.58–2.98)101.67 (0.88–3.15)1.56 (0.83–2.96)
 16–191092.61 (2.03–3.34)2.41 (1.87–3.11)431.62 (1.16–2.26)1.63 (1.17–2.28)
 ≥20552.01 (1.48–2.73)1.85 (1.36–2.53)521.50 (1.11–2.03)1.49 (1.10–2.01)
 p trend3 0.03060.0747 0.74040.6715
Pack-years
 ≤19211.71 (1.08–2.70)1.42 (0.89–2.26)261.61 (1.07–2.42)1.56 (1.04–2.35)
 20–39492.20 (1.59–3.05)1.97 (1.42–2.74)261.24 (0.82–1.87)1.26 (0.83–1.90)
 40–59742.85 (2.15–3.77)2.69 (2.02–3.58)161.16 (0.70–1.92)1.20 (0.72–2.00)
 ≥60572.14 (1.57–2.90)2.08 (1.52–2.84)202.73 (1.73–4.31)2.84 (1.79–4.51)
 p trend3 0.76130.4011 0.10100.0528

Among men who quit smoking cigarettes (Table IV), stomach cancer death rates consistently and significantly increased with smoking duration (p-trend = 0.002), cigarettes smoked per day (p-trend = 0.064) and pack-years of smoking (p-trend = 0.004) and decreased with years since smoking cessation (p-trend = 0.002); these trends were not seen in women who quit smoking. Although the proportion of men and women who had smoked more than 1 pack of cigarettes per day decreased with more years since quit smoking, results for smoking duration and time since cessation were not substantially different when controlling for smoking intensity (data not shown).

Table IV. Dose-Response Gradients Among Former Smokers of Exclusively Cigarettes1
 MenWomen
nAge-adjusted RR (95% CI)Multivariate-adjusted RR (95% CI)2nAge-adjusted RR (95% CI)Multivariate-adjusted RR (95% CI)2
  • 1

    Number of stomach cancer deaths (n), age- and multivariate-adjusted rate ratios (RR) and 95% confidence intervals (CI), CPS II men and women, 1982 through 1996.

  • 2

    Multivariate models include age, race, education, family history of stomach cancer, consumption of high-fiber grain foods, vegetables, citrus fruits or juices, and use of vitamin C, multivitamins, and aspirin.

  • 3

    Estimates of p trend excluded non-users of tobacco.

  • 4

    Two highest categories in women grouped together due to small numbers.

Never smoker1691.00 (Referent)1.00 (Referent)2821.00 (Referent)1.00 (Referent)
Cigarette smokers2851.56 (1.29–1.89)1.55 (1.28–1.88)1061.26 (1.01–1.59)1.36 (1.08–1.71)
Years smoked cigarettes
 <20571.07 (0.79–1.45)1.10 (0.81–1.50)361.25 (0.88–1.78)1.37 (0.96–1.96)
 20–29731.52 (1.15–2.01)1.54 (1.17–2.03)271.56 (1.05–2.33)1.72 (1.15–2.57)
 30–39741.73 (1.32–2.28)1.71 (1.30–2.26)281.10 (0.74–1.62)41.19 (0.81–1.77)4
 ≥40622.10 (1.56–2.83)2.03 (1.50–2.74)   
 p trend3 0.00050.0017 0.34990.3208
Cigarettes smoked per day
 <20601.32 (0.98–1.77)1.29 (0.96–1.73)511.11 (0.82–1.50)1.19 (0.88–1.62)
 20 (1 pack)971.60 (1.25–2.06)1.60 (1.24–2.06)271.43 (0.96–2.12)1.52 (1.02–2.27)
 21–39421.56 (1.11–2.19)1.61 (1.14–2.26)171.43 (0.87–2.34)41.55 (0.95–2.55)4
 ≥40 (≥2 Packs)761.84 (1.40–2.41)1.84 (1.40–2.42)   
 p trend3 0.07680.0644 0.19460.1651
Age started smoking cigarettes, years
 ≤15441.34 (0.96–1.86)1.27 (0.91–1.78)50.85 (0.35–2.06)0.88 (0.36–2.14)
 16–191331.56 (1.24–1.97)1.55 (1.23–1.96)401.17 (0.83–1.64)1.28 (0.91–1.80)
 ≥201051.67 (1.31–2.13)1.70 (1.33–2.17)591.41 (1.06–1.86)1.51 (1.14–2.01)
 p trend3 0.87700.6079 0.48720.6052
Pack-years
 ≤19641.13 (0.85–1.51)1.14 (0.86–1.53)501.29 (0.95–1.74)1.40 (1.03–1.91)
 20–39841.62 (1.25–2.11)1.63 (1.25–2.13)191.25 (0.78–1.99)1.36 (0.85–2.18)
 40–59521.86 (1.36–2.55)1.87 (1.37–2.56)121.12 (0.63–2.00)41.21 (0.67–2.16)4
 ≥60561.95 (1.44–2.64)1.94 (1.43–2.64)   
 p trend3 0.00320.0037 0.83610.8279
Age quit smoking cigarettes
 ≤30281.13 (0.75–1.69)1.15 (0.77–1.72)151.01 (0.60–1.72)1.11 (0.65–1.89)
 31–40541.23 (0.90–1.67)1.25 (0.92–1.70)291.69 (1.14–2.49)1.83 (1.24–2.72)
 41–50841.62 (1.25–2.11)1.62 (1.24–2.10)231.18 (0.77–1.82)1.27 (0.83–1.95)
 ≥511151.94 (1.52–2.47)1.88 (1.47–2.40)361.20 (0.85–1.71)1.28 (0.90–1.82)
 p trend3 0.00060.0015 0.84640.6833
Number of years since smoking cessation
 ≤101051.99 (1.56–2.55)1.92 (1.50–2.47)341.25 (0.87–1.79)1.31 (0.91–1.87)
 11–19861.65 (1.27–2.15)1.64 (1.26–2.14)311.35 (0.93–1.97)1.46 (1.00–2.13)
 ≥20901.21 (0.93–1.56)1.23 (0.95–1.59)381.22 (0.87–1.71)1.34 (0.95–1.89)
 p trend3 0.00060.0015 0.84640.6833

Current cigar smokers had significantly higher stomach cancer death rates than non-users of tobacco, with risk consistently increasing with smoking duration and number of cigars smoked per day (Table V). Smokers of 5 or more cigars per day (RR = 4.20, 95% CI = 2.32–7.60) and those who inhaled (RR = 3.93, 95% CI = 1.92–8.04) were at particularly high risk. There were too few stomach cancer deaths among former cigar smokers to conduct detailed analyses in that group.

Table V. Dose-Response Gradients Among Current Smokers of Exclusively Cigars1
 nAge-adjusted RR (95% CI)Multivariate-adjusted RR (95% CI)2
  • 1

    Number of stomach cancer deaths (n), age- and multivariate-adjusted ratios (RR) and 95% confidence intervals (CI), CPS II men, 1982 through 1996.

  • 2

    Multivariate models include age, race, education, family history of stomach cancer, consumption of high-fiber grain foods, vegetables, citrus fruits or juices, and use of vitamin C, multivitamins, and aspirin.

  • 3

    Estimates of p trend excluded non-users of tobacco.

Never user1691.00 (Referent)1.00 (Referent)
Cigar smokers252.47 (1.62–3.78)2.29 (1.49–3.51)
Years of smoking cigars
 ≤39132.54 (1.43–4.49)2.42 (1.36–4.28)
 ≥40102.77 (1.43–5.36)2.55 (1.32–4.94)
 p trend3 0.07880.1020
Cigars smoked per day
 1–4131.79 (1.01–3.18)1.68 (0.95–2.97)
 ≥5124.57 (2.54–8.22)4.20 (2.32–7.60)
 p trend3 0.74560.5880
Age started smoking cigars, years
 ≤1962.78 (1.23–6.28)2.47 (1.09–5.61)
 20–29123.01 (1.65–5.49)2.85 (1.56–5.19)
 ≥3051.74 (0.71–4.24)1.62 (0.67–3.97)
 p trend3 0.61650.5889
Inhalation
 Do not inhale152.21 (1.29–3.78)2.08 (1.22–3.57)
 Inhale slightly-deeply84.42 (2.17–9.00)3.93 (1.92–8.04)

Use of more than one type of tobacco

Men who reported currently using more than one type of tobacco had significantly higher stomach cancer death rates than non-users of tobacco and RR's increased with the number of years between first and last use of any tobacco product; the multivariate RR (95% CI) was 1.90 (1.20–3.02) for 30–39 years and 1.80 (1.33–2.43) for 40 years or more between first and last use. Men who had quit using any tobacco product at study enrollment were also at substantially increased risk with 40 or more years between first and last use of any tobacco (RR = 2.17, 95% CI = 1.47–3.20).

Tobacco use stratified by covariates, effect modification

Stomach cancer death rate ratios associated with cigarette or cigar smoking, or with use of more than one type of tobacco, were significantly higher (test for homogeneity p-value < 0.05) among men reporting a personal history of chronic indigestion or gastroduodenal ulcer than men without such history and higher among current aspirin users than non-users (Table VI). The increase in stomach cancer mortality associated with smoking duration was significantly greater among cigarette smoking men with a personal history of chronic indigestion or gastroduodenal ulcer than men without such history and higher among current aspirin users than non-users (data not shown). No effect modification was observed with consumption of vegetables, citrus fruits or juices, or alcohol in either gender.

Table VI. Tobacco Use Stratified by Selected Covariates1
 MenWomen
Current cigarette smokersFormer Cigarette smokersCurrent cigar smokersCurrent users of >1 type of tobaccoCurrent cigarette smokersFormer cigarette smokers
nRR (95% CI)nRR (95% CI)nRR (95% CI)nRR (95% CI)nRR (95% CI)nRR (95% CI)
  • 1

    Number of stomach cancer deaths (n), multivariate-adjusted rate ratios (RR) and 95% confidence interval (CI), CPS II men and women, 1982 through 1996. Multivariate models include age, race, education, family history of stomach cancer, consumption of high-fiber grain foods, vegetables, citrus fruits or juices, and use of vitamin C, multivitamins, and aspirin.

  • 2

    p-value for test of homogeneity across strata based on the Wald statistic, only p < 0.05 are noted.

  • 3

    Excluding persons with no information on birthplace.

Never user1691.00 (Referent)1691.00 (Referent)1691.00 (Referent)1691.00 (Referent)2821.00 (Referent)2821.00 (Referent)
Tobacco use2302.16 (1.75–2.67)2851.55 (1.28–1.88)252.29 (1.49–3.51)961.81 (1.40–2.35)1051.49 (1.18–1.88)1061.36 (1.08–1.71)
Personal history of chronic indigestion or gastroduodenal ulcer
 None1671.92 (1.52–2.42)2091.36 (1.10–1.68)161.59 (0.94–2.69)671.52 (1.13–2.04)891.47 (1.15–1.90)971.44 (1.13–1.83)
 Yes633.45 (2.05–5.80)762.62 (1.58–4.34)98.93 (4.02–19.9)293.45 (1.93–6.20)161.56 (0.85–2.86)90.87 (0.42–1.83)
 p homogeneity2 0.0419 0.0182 0.0003 0.0141    
Current aspirin use
 No1241.69 (1.27–2.25)1541.25 (0.96–1.63)102.23 (1.28–3.88)561.30 (0.90–1.89)471.60 (1.17–2.18)541.39 (1.01–1.92)
 Yes1062.79 (2.06–3.77)1311.95 (1.47–2.60)152.35 (1.21–4.57)402.50 (1.75–3.57)581.38 (0.98–1.93)521.32 (0.96–1.82)
 p homogeneity2 0.0187 0.0238   0.0132    
Family history of stomach cancer
 No2172.11 (1.69–2.62)2711.54 (1.26–1.87)232.18 (1.40–3.40)871.71 (1.31–2.24)991.50 (1.18–1.91)1001.38 (1.09–1.75)
 Yes133.53 (1.40–8.89)141.82 (0.74–4.53)24.92 (1.01–23.9)94.27 (1.58–11.5)61.35 (0.54–3.35)61.12 (0.45–2.79)
Own or parents' birthplace3
 US or Canada1732.50 (1.93–3.23)2021.82 (1.43–2.32)111.74 (0.92–3.26)752.13 (1.57–2.90)721.39 (1.05–1.83)791.42 (1.09–1.86)
 Other521.56 (1.07–2.26)791.13 (0.81–1.57)132.70 (1.47–4.94)201.38 (0.83–2.29)331.76 (1.17–2.65)271.09 (0.70–1.69)
Alcohol
 Non-drinker942.12 (1.59–2.84)1291.60 (1.23–2.07)91.98 (1.00–3.93)381.55 (1.06–2.27)581.56 (1.16–2.11)591.50 (1.12–2.01)
 Drinker1362.10 (1.55–2.86)1561.48 (1.11–1.99)162.44 (1.40–4.24)581.99 (1.38–2.86)471.39 (0.95–2.04)471.19 (0.81–1.74)

Attributable fraction in the population

Assuming that the multivariate-adjusted RR's from our study represent a causal relationship with tobacco and using prevalence estimates of tobacco use from the 1991 National Health Interview Survey,4 we estimated the proportion of stomach cancer deaths in the US population attributable to ever use of any tobacco to be 28% in men and 14% in women, with the majority of deaths attributable to continuing use. In men, nearly 25% of stomach cancer deaths could be attributed to current cigarette smoking, over 5% to current cigar smoking and 4% to current smokeless tobacco use.

DISCUSSION

Results from this large prospective study strongly support a causal role for tobacco use in the development of stomach cancer. Stomach cancer death rates were consistently highest in men and women who currently used any tobacco product, intermediate in former users and lowest in never-users. Significant increases in risk were associated with use of different tobacco products in men and with cigarette smoking in women. The magnitude of the increased risk was similar in men who smoked cigarettes or cigars and increased with smoking duration and intensity. CPS II is the first US prospective study to report a positive association between cigarette smoking and stomach cancer death in women.

Three prospective studies28, 58, 59 of women in Scandinavia reported no association between cigarette smoking and stomach cancer mortality or incidence. All 3 cohorts, however, combined yielded a sample size less than 20% of the CPS II female cohort and women in one of these studies28 were substantially younger than those in CPS II. Case-control studies21, 33, 35, 38, 41, 42, 46, 48, 50 have generally reported risk estimates for both sexes combined. Only 5 case-control studies37, 40, 47, 51, 52 have reported risk estimates separately for women; 3 such studies published in the 1990s from the US,40 Japan,51 and Poland52 reported significantly increased risk associated with current cigarette smoking.

Few studies have reported on cigar, pipe, or smokeless tobacco use in relation to stomach cancer and not all of these studies have excluded users of other tobacco products from the referent group. In CPS II, current smokers of exclusively cigars, even men who reported no inhalation, had significantly higher stomach cancer mortality than non-users of any tobacco. The lack of a significant trend associated with cigar smoking duration may relate to the fact that cigar smoking may be initiated at any age throughout adulthood,77 whereas cigarette smoking begins primarily before age 25 years. The comparable magnitude of increased risk of stomach cancer death in current smokers of cigars or cigarettes resembles the relationship between these products and oral and esophageal cancers,78 reflecting the tendency of cigar smokers to swallow particles of tobacco leaf and dissolved constituents of smoke, thus directly exposing oral and gastrointestinal tissues to tobacco carcinogens.78

Snuff and chewing tobacco also can be swallowed with saliva and come in direct contact with gastric mucosa. Although 3 Swedish studies42, 50, 55 have reported no association between snuff dipping and stomach cancer incidence, only 1 study50 explicitly excluded users of other tobacco products from its referent group. Snuff production in Sweden involves heat processing rather than curing and fermentation as is done in the US.55 Fermentation increases N-nitrosation of nicotine and may account for the markedly higher (up to 250-fold) amounts of tobacco-specific N-nitrosamines and volatile N-nitrosamines found in US moist snuff compared to European snuff.79, 80 Levels of hemoglobin adducts of tobacco-specific N-nitrosamine metabolites have been shown to be higher in US snuff dippers than even tobacco smokers.81

The strength of association between cigarette smoking and stomach cancer has been stronger in studies published since 1990 than in studies published previously. Earlier studies showed relative risk estimates of approximately 1.5,22, 23, 24, 33, 34, 35 whereas the majority of prospective25, 26, 27, 28, 31 and case-control studies21, 36, 37, 40, 41, 42, 43, 45, 46, 49, 51, 52, 53, 54 published since 1990 report current cigarette smoking risk estimates of 2.0 or greater. Current cigarette smoking risk estimates in case-control studies21, 40, 45, 53, 56 have also been consistently higher for gastric cardia than non-cardia adenocarcinomas.

Several studies,21, 25, 29, 31, 42, 48, 49, 53 including CPS II, show that the relative risk estimates associated with current smoking steadily increases with the number of cigarettes smoked per day up to the highest exposure category, but decreases in the highest category. This lack of a sustained linear trend in risk with smoking intensity may relate in part to the decreased tolerance of people who develop chronic atrophic gastritis, intestinal metaplasia, or gastroduodenal ulcers to continued or heavy smoking. Risk estimates increased consistently with smoking intensity among CPS II men who had quit smoking cigarettes (p-trend = 0.002) and were markedly higher among former smokers who reported a history of chronic indigestion or gastroduodenal ulcers. The inconsistent dose-response relationship in the literature also may reflect the topographic and histologic heterogeneity of stomach cancer cases in different studies; the association with tobacco use may differ by tumor type. Nevertheless, CPS II and nearly all case-control studies published since 199021, 42, 47, 48, 49, 50, 52, 53, 56 show increasing risk of stomach cancer with cigarette smoking duration. CPS II data in men further show significantly decreasing trend in risk with more years since smoking cessation.

CPS II is the only prospective study that has adjusted risk estimates for potential confounders including education, consumption of fruits and vegetables and intake of supplemental vitamin C and multivitamins. Adjustment for these covariates had little effect on the significantly increased risk estimates associated with tobacco use. One study82 has adjusted directly for H. pylori infection; this study reported a statistically significant odds ratio of 2.2 for current cigarette smoking, nearly identical to the results of CPS II and other studies unadjusted for H. pylori infection.

The association between stomach cancer death and cigarette or cigar smoking in CPS II men was significantly modified by a personal history of chronic indigestion or gastroduodenal ulcer. Because these conditions are strongly associated with chronic and recurrent H. pylori infection,14, 16 we hypothesize that H. pylori and tobacco smoke may act together to increase the risk of stomach cancer. Nicotine has been shown to potentiate the effect of H. pylori by increasing vacuolating toxin activity in cells.83 Among patients with H. pylori-associated gastritis, smokers have greater gastric chemokine mRNA expression, leading to more severe gastritis, than non-smokers.84 Smokers with chronic atrophic gastritis are consistently more likely than non-smokers to progress to intestinal metaplasia or dysplasia.85, 86, 87, 88 Although we had no data on H. pylori infection, 2 case-control studies53, 82 have reported higher cigarette smoking OR among H. pylori infected than non-infected gastric cancer cases. Our finding of a significantly stronger association between cigarette smoking and stomach cancer death in current aspirin users suggests that tobacco smoke and nicotine may exacerbate existing gastric mucosal lesions or interfere with wound healing response of the gastric mucosa to injury.89

It is biologically plausible that tobacco products play a causal role in gastric carcinogenesis at both early and late stages, possibly with decades between exposure and clinical disease. Tobacco products contain a myriad of carcinogens associated with gastric adenocarcinoma in experimental animals and humans.79, 90, 91 These carcinogens have direct contact with stomach mucosa when swallowed with saliva or bronchial secretions, as well as indirect exposure via the bloodstream. Evidence that tobacco carcinogens can bind to gastric mucosal DNA come from studies showing smokers to have significantly higher levels of smoking-related DNA adducts than non-smokers,92 and significantly higher levels of benzo(a)pyrene-DNA adducts in gastric tumors compared to normal tissue.93 Tobacco smoking may promote stomach carcinogenesis by causing mucosal injury through increasing gastric acid and pepsin secretion94, 95 or reflux of duodenal bile salts,96 or by depleting the concentrations of antioxidants in gastric juice or plasma.97, 98 Further research is needed to clarify the role of tobacco smoking and tobacco carcinogens in the complex web of dietary, infectious and host susceptibility factors important in stomach cancers of distinct topographic and histologic subtypes.

Our study was unable to differentiate between death due to tumors of the gastric cardia or non-cardia. Assuming that the distribution of stomach cancers by anatomical subsite in CPS II is similar to that in the US general population, we expect that half of the stomach cancer deaths in men and a smaller proportion in women resulted from tumors of the gastric cardia.17 Based on the consistently higher cigarette smoking risk estimates for gastric cardia than non-cardia tumors in case-control studies, CPS II results likely underestimated the true association between tobacco use and gastric cardia tumors. CPS II estimates also may have underestimated the true risk associated with continuing tobacco use since some current users at study enrollment may have quit over the follow-up period. The lack of updated tobacco use information during the follow-up period also weakens our ability to evaluate dose-response by smoking intensity or duration because current heavy smokers at study enrollment may discontinue or reduce tobacco consumption over time. The strengths of our study lie in the detailed information on the use of multiple types of tobacco products and covariates and on its large size that allowed us to examine the exclusive use of different tobacco products.

Worldwide, it has been estimated66 that the smoking-attributable proportion of stomach cancers is 11% among men and 4% among women in developing countries and 17% among men and 11% among women in developed countries. Based on CPS II results, we estimate that 28% of stomach cancer deaths in US men and 14% in women are attributable to tobacco use. The proportion of incident gastric cancers attributable to tobacco smoking has been estimated to be 20% in Poland52 and 31% in India.99 Adding stomach cancer to the list of cancers caused by tobacco would increase the total number of smoking-attributable deaths by at least 84,000 per year worldwide,66 not accounting for recent increases in smoking prevalence or the expected rise in stomach cancer incidence due to aging in the developing world.6

Tobacco use, particularly cigarette and cigar smoking, is associated with increased stomach cancer mortality in this large cohort of US adults. Results from this and other studies are consistent with a causal role of prolonged tobacco use in stomach cancer incidence and mortality in men and women and support reconsidering stomach cancer as a tobacco-related cancer.

POSTSCRIPT

Stomach cancer was upgraded to a cancer for which sufficient evidence exists to infer a causal relationship with tobacco by the International Agency for Research on Cancer in its Review of Tobacco conducted in June 2002.

Acknowledgements

Cancer Prevention Study II is funded by the American Cancer Society. A.C. is supported by the American Cancer Society and Public Health Service grant K07CA75062 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services.

Ancillary