Smokeless tobacco use and risk of cancer of the pancreas and other organs
Limited data are available on the carcinogenicity of smokeless tobacco products in organs other than the mouth. Snus is a smokeless tobacco product widely used in Norway. We studied 10,136 Norwegian men enrolled since 1966 in a prospective cohort study, 31.7% of whom were exposed to snus. The relative risk of pancreatic cancer for snus use was 1.67 (95% confidence interval [CI] = 1.12, 2.50); that of oral and pharyngeal cancer was 1.10 (95% CI = 0.50, 2.41), that of esophageal cancer was 1.40 (95% CI = 0.61, 3.24), and that of stomach cancer was 1.11 (95% CI = 0.83, 1.48). The relative risks of cancers of the lung (either all histological types or adenocarcinoma), urinary bladder and kidney were not increased among snus users. The increase in the relative risk of pancreatic cancer was similar in former and current snus users and was restricted to current tobacco smokers. Our study suggests that smokeless tobacco products may be carcinogenic on the pancreas. Tobacco-specific N-nitrosamines are plausible candidates for the carcinogenicity of smokeless tobacco products in the pancreas. © 2004 Wiley-Liss, Inc.
Use of smokeless tobacco products is common in many regions of the world and is increasing in the United States and Northern Europe.1, 2 Tobacco chewing is a major risk factor for oral and pharyngeal cancer in Asia,3, 4 but a similar increase in risk has not been shown consistently among users of smokeless tobacco products in the United States or Europe.5, 6 Smokeless tobacco might cause other cancers, in particular those linked to tobacco smoking, but limited data are available.2, 4, 5 In particular, an increased risk of pancreatic cancer has been suggested in studies based on few exposed cases.7, 8 A detailed assessment of the risk of pancreatic and other cancers entailed by smokeless tobacco use is needed before conclusions on the overall health risks of this group of products can be reached.
Although tobacco snuff and chewing entail very little exposure to polycylic aromatic compounds, exposure to N-nitrosamines is substantial.9, 10, 11 Tobacco-specific nitrosamines are experimental carcinogens and are heavily suspected to cause cancer, in particular adenocarcinoma, in humans.12
Snus is a smokeless tobacco product widely used in Norway; it is usually placed behind the upper or lower lip. The average sale of snus in the mid-1960s was around 200 g/year/Norwegian adult. It decreased to 80 g/year in the 1980s, and has remained stable since.13
We conducted a detailed analysis of cancer incidence in a cohort of Norwegian men to estimate the risk of cancer of the pancreas and other organs from use of smokeless tobacco products.
Material and methods
The cohort under study consists of 2 groups of subjects: a systematic sample of the general adult population of Norway identified from the 1960 census, and relatives of Norwegian migrants to the United States.7, 14 Study subjects completed questionnaires on lifestyle habits in 1964 and 1967. The participation rate varied by study and location, but was above 75%. The questionnaire collected information on use of smokeless tobacco, as well as information on dietary habits, tobacco smoking, alcohol drinking and anthropometric parameters.
A total of 12,431 men who were alive on 1 January 1966 were included in the study. Information on snus use was missing for 2,295 of them (18.5%); the remaining 10,136 cohort members were classified as regular current users (N = 1,999, 19.7%), regular former users (N = 1,216, 12.0%), or never or occasional users (N = 6,921, 68.3%). The age distribution of subjects with information on snus was very close to that of subjects with missing information (χ2 test with 7 d.f., p = 1.0). Tobacco smoking was classified as never/current/former smoking of cigarettes/cigars/pipe. Amount of current smoking was classified in 3 categories for cigarettes (1–9, 10–14 and 15+ cigarettes/day) and in 2 categories for cigars and pipe (1–4 and 5+ g/day). Information on amount of smoking was not available for former smokers. No reassessment of snus use or tobacco smoking was carried out during the follow-up.
Cohort members were followed until date of diagnosis of cancer, date of emigration, date of death or 31 December 2001, whichever occurred earliest. The follow-up was carried out via linkage with nationwide residence, mortality and cancer incidence registries, using unique personal identification numbers. Fifteen cohort members were lost to follow-up (0.15%). For the purpose of this analysis, we considered the incidence of cancers of the oral cavity and pharynx (ICD7, 141–148), esophagus (ICD7, 150), stomach (ICD7, 151), pancreas (ICD7, 157), lung (ICD7, 162), kidney (ICD7, 180) and urinary bladder (ICD7, 181). In addition, cases of esophageal and lung adenocarcinoma were considered separately. Cases diagnosed on the basis of a clinical examination or death certificate only were excluded. The analysis of pancreatic cancer risk was based on 220,007 person-years of observation. Censoring the follow-up at the time of first diagnosis resulted in a slightly different number of person-years in the analysis of each cancer.
Cox proportionate hazard regression models, including attained age as time variable, were fitted to the data to estimate relative risks (RR) and 95% confidence intervals (CI) of each cancer. The regression models used in the main analysis included terms for never, former and current smoking of cigarettes, cigars and pipe. In sensitivity analyses, alternative approaches were used to control for the potential confounding effect of tobacco smoking. Additional models included a term for body mass index (BMI).
The number of incident cases was 34 for oral and pharyngeal cancer, 27 for esophageal cancer (4 cases of adenocarcinoma), 217 for stomach cancer, 105 for pancreatic cancer, 343 for lung cancer (of whom 50 cases of adenocarcinoma), 88 for kidney cancer and 238 for bladder cancer.
There was an increased RR of pancreatic cancer among ever users of snus, and the RR of oral and pharyngeal, esophageal and stomach cancer showed a modest, non-significant increase. There was no increase in the RR of lung cancer (all histological types and adenocarcinoma) and of other cancers included in the analysis (Table I). No difference in the RR of pancreatic cancer was evident between former and current use. The number of cases of esophageal adenocarcinoma was too small to justify a separate analysis. Table II shows the RRs of pancreatic and lung cancers for ever snus use, estimated separately according to smoking habit. The number of cases among never and former smokers was small, and there was no evidence of an increased RR of pancreatic cancer in these 2 groups.
Table I. RR of Selected Cancers for Use of Smokeless Tobacco (snus)
|Lung cancer (all types)||271||72||0.80||0.61–1.05||28||0.80||0.54–1.19||44||0.80||0.58–1.11|
Table II. RR of Pancreatic Cancer and Lung Cancer for Ever Use of Smokeless Tobacco (Snus), Estimated Separately According to Smoking Status
Different approaches to control for the potential confounding effect of tobacco smoking resulted in risk estimates that were similar to those reported in Table I. For example, the RR of pancreatic cancer for ever snus use, derived from a model including a continuous term for amount of tobacco smoking, was 1.66 (95% CI = 1.06, 2.62). Further adjustment for body mass index did not affect the RR (not shown in detail).
Our study provides evidence for a carcinogenic effect of smokeless tobacco products on the pancreas, thus confirming the findings of an early report from this cohort, which was based on only 14 cases among snus users.7 None of the other available studies, all from the United States, included more than 10 cases of pancreatic cancer among users of smokeless tobacco products.8, 15, 16, 17 Despite the low statistical power, an association was suggested in two of these studies.8, 17
Arguments in favor of a causal effect of snus on pancreatic cancer in our study are the strong statistical significance, the likely exclusion of selection and information bias because of the prospective nature of the investigation, and the lack of an apparent confounding effect of tobacco smoking and BMI. Residual confounding by tobacco smoking or by other potential risk factors for pancreatic cancer, such as heavy alcohol intake and a diet poor in fruits and vegetables, cannot be completely ruled out. The lack of a corresponding increase in risk of lung cancer detracts from the hypothesis of residual confounding by tobacco smoking.
Lack of information on snus use and tobacco smoking after enrollment in the cohort is a matter of concern, in particular given the long-term follow-up of the study. Given the decrease in the prevalence and use of snus among Norwegian men during the study period, it is likely that change in snus use status mainly affected current users who quit rather than non-users and formed users who took up the habit. Because misclassification is unlikely to have occurred differentially with respect to outcome (i.e., future cases of pancreatic cancer having changed their habits during the follow-up differently from other cohort members), it should have resulted in an underestimate of the difference of carcinogenic effect of snus between current and former snus users. Additional limitations of our study are the lack of information on amount and duration of snus use, which preempted dose-response analyses, and the small number of cases of pancreatic cancer among never and former smokers.
N-nitroso compounds, specifically N-nitrosamines, are plausible candidates for the carcinogenicity of smokeless tobacco products in the pancreas. Tobacco-specific nitrosamines have been identified in the pancreatic juice of smokers and, to a lesser extent, of non-smokers.18 Experimental studies have shown the ability of tobacco-specific nitrosamines to produce pancreatic cancer in exposed rats,12 and, in one experiment, oral administration of NNK (one of the main tobacco-specific nitrosamines) was more effective in causing pancreatic cancer than other routes of exposure.19 Furthermore, a high proportion of G to A transitions in K-ras mutations detected in nitrosamine-induced animal pancreatic cancers represents further evidence for a central role of tobacco-specific nitrosamines and other N-nitrosamines in pancreatic carcinogenesis, although results on mutations in human cancers are not consistent.12, 20
The lack of an increased risk of lung cancer among smokeless tobacco users confirms previous reports.21, 22, 23 The relatively large size of the cohort confers a power of 80% to detect as significant a relative risk of 1.28 or greater. The analysis of lung adenocarcinoma was limited by the small number of cases, however, and our study had 80% power to detect as significant a relative risk of 1.85 or greater. We therefore cannot exclude some carcinogenic effect of smokeless tobacco on lung adenocarcinoma.
A weak, non-significant association was detected between use of snus and cancer of the oral cavity. The statistical power of our study for oral cancer was similar to that for lung adenocarcinoma (80% power to detect as significant a relative risk of 1.8–1.9). Chewing of tobacco products is an important cause of oral and pharyngeal cancer in several developing regions of the world, including in particular India,3, 9 other South Asian countries such as Pakistan, Bangladesh and Myanmar,24, 25, 26 Central Asia,27 and Sudan.28, 29 Studies conducted in the United States provided evidence for a carcinogenic effect of smokeless tobacco products on the oral cavity,30, 31, 32, 33 although these findings were not confirmed by other investigations conducted in the United States34, 35 or in Sweden.36, 37 The inconsistencies of results from previous studies can be explained by methodological aspects such as adequacy of control for tobacco smoking and statistical power; in any case, our results are consistent with previous evidence in supporting the conclusion that it is unlikely that the use of smokeless tobacco products in Europe and United States entails a substantial increase in the risk of oral and pharyngeal cancer. The reasons for the difference in carcinogenic risk entailed by smokeless tobacco products used in Europe, as compared to those used in developing countries, are not fully understood, but they might be related to tobacco species, fermentation and ageing.38
No effect of snus use on esophageal36, 39, 40 and stomach cancer41 was detected in previous studies, and our results might be attributed to chance. Previous studies of cancers of the bladder42, 43 and kidney44, 45 do not suggest an association with use of smokeless tobacco products, which is in agreement with our findings.
There is controversy on whether the use of smokeless tobacco products that are common in Northern Europe should be encouraged as an alternative to tobacco smoking, due to the apparent lack of a strong carcinogenic effect on organs such as the lung and the oral cavity.3, 46 Although the risk of cancer of the lung and some other organs in this population was lower among snus users than among non-users, the decrease was of small magnitude and not statistically significant, and there was no clear evidence of a beneficial effect among non-smokers. Our study does not offer arguments in favor of the use of smokeless tobacco products to reduce the burden of tobacco-related cancer incidence or mortality. Furthermore, it provides evidence of a carcinogenic effect on the pancreas, which should be taken into account in the assessment of the health effects of this group of products.