American Journal of Epidemiology Advance Access originally published online on January 22, 2007
American Journal of Epidemiology 2007 165(7):814-820; doi:10.1093/aje/kwk066
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ORIGINAL CONTRIBUTIONS |
Contribution of Tobacco and Alcohol to the High Rates of Squamous Cell Carcinoma of the Supraglottis and Glottis in Central Europe
1 International Agency for Research on Cancer, Lyon, France
2 Cancer Research Centre, Moscow, Russia
3 Institute of Occupational Medicine, Lodz, Poland
4 Institute of Hygiene, Public Health, Health Services, and Management, Bucharest, Romania
5 Specialized State Health Institute, Banská Bystrica, Slovakia
6 National Institute of Environmental Health, Budapest, Hungary
Correspondence to Dr. Paul Brennan, Genetic Epidemiology Group, International Agency for Research on Cancer, 150 cours Albert Thomas, 69008 Lyon, France (e-mail: brennan{at}iarc.fr).
Received for publication August 9, 2005. Accepted for publication September 18, 2006.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Incidence rates for laryngeal cancer in Central Europe are among the highest in the world. The authors recruited cases and controls between 2000 and 2002 for the Central and Eastern Europe Multicenter Study to investigate the role of tobacco and alcohol as causes of laryngeal cancer in this region. A total of 384 incident squamous cell cases were included, comprising 254 glottic and 108 supraglottic cancers. Hospital controls were chosen from within the same catchment area, from diseases unrelated to tobacco or alcohol (n = 918). Significant dose-response trends for frequency and duration of tobacco use were observed for both supraglottic and glottic cancers, with potentially a more important effect for supraglottic cancer. Quitting smoking was protective against laryngeal cancers after 5 years. Any increases in risk for alcohol drinking were generally moderate and nonsignificant. A greater than multiplicative interaction was observed between tobacco and alcohol on the risk of laryngeal cancer (p = 0.04). Approximately 87% of laryngeal cancer cases in Central Europe are attributable to tobacco use, of which 75% and 12% are due to current and past smoking, respectively. Approximately 39% are attributable to the interaction between alcohol and tobacco. Preventive efforts to encourage current smokers to quit are likely to be the most effective way to reduce the incidence of laryngeal cancer in this region.
alcohol drinking; carcinoma, squamous cell; Europe; glottis; laryngeal neoplasms; tobacco
Abbreviations: CI, confidence interval; ICD-10, International Classification of Diseases, Tenth Revision; OR, odds ratio
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
In the year 2000, approximately 23,000 incident cases of laryngeal cancer and 14,000 deaths due to laryngeal cancer are estimated to have occurred in Central and Eastern Europe (1). On a regional basis, the age-adjusted incidence rate of laryngeal cancer for men from Central and Eastern Europe is the highest in the world. Incidence rates are 13.1 per 100,000 in Hungary, 12.5 per 100,000 in Russia, and 12.1 per 100,000 in Poland. In comparison, the incidence rates for women are very low, with an overall rate of 0.5 per 100,000 for Central and Eastern Europe. Mortality rates for tobacco-related cancers have increased substantially over recent decades and are projected to increase further for 2005–2009, particularly for laryngeal cancers among Hungarian men (2).
The main risk factors for laryngeal cancer in Western countries are known to be tobacco use and alcohol consumption, with a multiplicative effect between the two (3, 4). It is important to explore the role of tobacco and alcohol for laryngeal cancer in Central and Eastern Europe and to assess whether differences exist in risk factors for supraglottic and glottic tumors, in order to understand the reasons for the very high incidence and to inform public health efforts. To this end, we report on a multicenter case-control study on laryngeal cancers in Central Europe.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
A multicenter case-control study was conducted in the following five centers in Central and Eastern Europe: Bucharest (Romania), Budapest (Hungary), Lodz (Poland), Moscow (Russia), and Banksa Bystrika (Slovakia). The overall recruitment period at these centers was from 2000 to 2002. Of the 397 laryngeal tumor cases ascertained (International Classification of Diseases, Tenth Revision (ICD-10), codes C32.0–32.9), 13 subjects were excluded because the histologic type was missing, was not squamous cell carcinoma, or was an in situ carcinoma. Thus, a total of 384 laryngeal squamous cell carcinoma cases, diagnosed at designated hospitals or cancer clinics and confirmed histologically or cytologically, were recruited into the study within 3 months of diagnosis. They included 254 squamous cell carcinoma cases of the glottis (ICD-10 code C32.0) and 108 squamous cell carcinoma cases of the supraglottis (ICD-10 code C32.1), the remaining cases being subglottis (ICD-10 code C32.2), overlapping (ICD-10 code C32.8), or not specified (ICD-10 code C32.9).
Controls were frequency matched from inpatients or outpatients in the same hospital as the cases. Only controls with a recent diagnosis from a defined list of diseases unrelated to tobacco and alcohol were included. The distribution of control diseases was as follows: genitourinary, 10 percent; eye, 10 percent; the musculosketal system, 9 percent; skin, 8 percent; circulatory system, 8 percent; gallbladder or biliary tract, 8 percent; other digestive, 8 percent; central nervous system, 7 percent; hernias, 5 percent; injury, 4 percent; metabolic, 2 percent; appendicitis, 2 percent; mental, 2 percent; respiratory, 2 percent; and a mixture of other diseases, 14 percent, none of which contributed more than 2 percent overall. In Moscow, the controls were frequency matched to the laryngeal cancer cases by age, sex, and referral or residence area. In the other centers, controls overlapped with those for the case-control study of lung cancer conducted according to an identical protocol (5, 6). Because the lung cancer study was started earlier than the laryngeal cancer study, we excluded controls if their interview date was more than 6 months earlier than the first interview date for the head and neck cancer cases. Written consent for participation was obtained from all study subjects, and ethical approval has been obtained from all study centers as well as for the International Agency for Research on Cancer. Cases and controls were interviewed with a structured questionnaire on residential and lifestyle history by the same team of interviewers in each center. The lifestyle questionnaire included basic demographic characteristics, history of tobacco use and alcohol consumption, and a food frequency questionnaire focusing on commonly consumed fruits and vegetables.
With regard to statistical analysis, unconditional logistic regression was used to estimate odds ratios and 95 percent confidence intervals for tobacco use and alcohol consumption with SAS, version 9.1, statistical software (SAS Institute, Inc., Cary, North Carolina) after adjustment for potential confounders such as age, sex, education, center, body mass index, fruit intake, vegetable intake, tobacco smoking, and alcohol drinking, where appropriate.
The smoking history section of the questionnaire was detailed, with information on age at which regular smoking began and stopped, periods in which the subject smoked by age, the number and type of cigarettes smoked during that period, and any period of not smoking. Overall tobacco use was estimated by combining the amount of cigarettes, cigars/cigarillos (one cigar/cigarillo = four cigarettes), and pipes (one pipe = 3.5 cigarettes) and quantified into the equivalent of cigarettes per day. Ever smokers were defined as having smoked at least 100 cigarettes in an entire lifetime or having regularly smoked cigars/cigarillos or pipes. Former smokers were defined as having quit smoking for at least 2 years, prior to interview. Because all of the supraglottic cancer cases were ever smokers, we used light smokers as the reference group for assessing frequency, duration, and pack-years for the three outcomes (glottis, supraglottis, overall), so that the risk estimates could be compared. Pack-years of smoking were calculated by multiplying the number of packs smoked per day (20 cigarettes in a pack) by the number of years the subject smoked.
Ever drinkers were defined as having ever consumed alcoholic drinks. Given the small number of never drinkers among cases, we used light drinkers as the reference group. Each subject was asked to recount his or her drinking of beer, wine, and spirits in a typical week during specific age periods. The grams of ethanol per week were subsequently estimated, assuming that beer is approximately 5 percent ethanol, wine is 12 percent ethanol, and spirits are 40 percent ethanol in volume (3). Cumulative consumption was estimated by multiplying grams per week by the years of alcohol consumption.
Body mass index was calculated from the subject's self-reported weight 2 years prior to the interview. Fruit and vegetable intakes were estimated from a 20-item food frequency section of the questionnaire, in the number of days per week. p values for dose-response relations were assessed by including the categorical variable in the logistic regression model.
Interactions between tobacco and alcohol were assessed by comparing the fit of a regression model including terms for use of tobacco alone, consumption of alcohol alone, and both with that of a model including also an interaction term between the two habits. Durations of tobacco and alcohol habits were also assessed for interaction effects. Finally, we calculated population attributable fractions for tobacco and alcohol use for laryngeal cancers, on the basis of the logistic regression models (7).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Selected characteristics of the laryngeal cancer cases and controls are presented in table 1. The largest number of laryngeal cancer cases was from Poland. The age distribution was similar between cases and controls, although more of the cases were men. Relative to the controls, the cases had fewer years of formal education, a lower body mass index, and lower fruit and vegetable intakes.
|
Current smokers were at an increased risk of laryngeal cancer (odds ratio (OR) = 2.88, 95 percent confidence interval (CI): 1.95, 4.24), while nonsmokers were protected against laryngeal cancer (OR = 0.19, 95 percent CI: 0.10, 0.40) relative to former smokers (table 2). When nonsmokers were taken as the reference group, we estimated a 15-fold increase in risk of laryngeal cancer among current smokers and a fivefold increase among former smokers. Analysis by subsite was complicated by the absence of nonsmoking cases of the supraglottis, and all subsequent analyses have used the lowest level of smokers as the reference category.
|
The increased risk for current smokers when compared with former smokers was similar for supraglottic and glottic cancers. Although the risk estimates in the upper categories of duration and pack-years of tobacco use were more elevated for supraglottic cancer, the confidence intervals overlapped with those of glottic cancer. Dose-response trends for frequency, duration, and pack-years of tobacco use were observed for the two subsites and overall laryngeal squamous cell carcinoma. A protective effect of time since quitting was apparent for both supraglottic and glottic cancers, with decreasing risk observed for longer time since quitting. After 20 years since quitting smoking, the decrease in risk approached that of never smokers. Elevated odds ratios for the use of cigarettes only, cigars only, and both cigarettes and cigars were observed for laryngeal cancers. The carcinogenic effect of pipe smoking was difficult to estimate because of the small number of pipe smokers.
For alcohol consumption, risk estimates for frequency, duration, and cumulative consumption categories were moderate for both glottic and supraglottic cancers (table 3). Essentially all of the supraglottic cancer patients were ever drinkers. Dose-response relations were suggested for the duration of drinking (ptrend = 0.06) and cumulative consumption (p = 0.01) on the risk of laryngeal squamous cell carcinoma overall. Differences in risk due to alcohol beverage type were not observed.
|
An interaction that was more than multiplicative was observed between ever tobacco use and alcohol use. The odds ratios were 1.71 (95 percent CI: 0.28, 10.41) for use of tobacco alone, 0.44 (95 percent CI: 0.11, 1.78) for use of alcohol alone, and 5.79 (95 percent CI: 1.65, 20.29) for use of both tobacco and alcohol (likelihood test p value = 0.04), after adjustment for age, sex, education, center, body mass index, fruit intake, and vegetable intake. When we further explored interactions between years of alcohol consumption and tobacco use (table 4), the highest odds ratio was observed for persons who had consumed alcohol and used tobacco for 40 or more years (OR = 14.37, 95 percent CI: 5.21, 39.65).
|
Population attributable fractions were 87 percent (95 percent CI: 76, 93) for tobacco use and 39 percent (95 percent CI: 0, 77) for alcohol use, when adjusted for potential confounders. Given the absence of an independent alcohol effect, all of the attributable fraction for alcohol use can be interpreted as being due to the interaction between alcohol and tobacco. Additionally, the attributable fractions were 12 percent (95 percent CI: 4, 20) for past smokers and 75 percent (95 percent CI: 68, 81) for current smokers. For supraglottic cancers, the population attributable fractions were 12 percent (95 percent CI: 0.0, 27) for past tobacco smokers and 88 percent (95 percent CI: 80, 92) for current tobacco smokers. For glottic cancer, the population attributable fractions were 11 percent (95 percent CI: 0, 21) for past tobacco smokers and 72 percent (95 percent CI: 61, 80) for current tobacco smokers.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The carcinogenic risk of tobacco smoking was recently reevaluated (4), and the average relative risk for laryngeal cancer across the 30 studies reviewed was estimated as tenfold (8). The magnitude of risk observed in our study is comparable, with a 15-fold increase in risk of laryngeal cancer among current smokers and a fivefold increase among former smokers. For alcohol drinking, our risk estimates were moderate, with an approximate increase in laryngeal cancer risk from 1.5-fold to 2.0-fold for the upper categories of cumulative consumption. Furthermore, we did not identify an independent effect of alcohol consumption but observed a greater than multiplicative effect between alcohol consumption and tobacco smoking on the risk of laryngeal cancer.
The only previous study in Central Europe (9) reported high relative risks for laryngeal cancer due to tobacco and alcohol, in Lower Silesia, Poland, including 249 laryngeal cancer cases and 965 controls. Zatonski et al. (9) reported odds ratios of 59.7 (95 percent CI: 13.0, 274) for smoking more than 30 cigarettes per day and 10.4 (95 percent CI: 4.0, 27.2) for drinking vodka for more than 30 years. The difference in risk estimates in the study by Zatonski et al. and our study may be explained by several differences between the two studies. Our study included both men and women, while the study by Zatonski et al. included men only. The latter study focused on one province in Poland, while our study included several countries within Central and Eastern Europe. It may be possible that there is a risk difference between the regions of Central and Eastern Europe. Although such stratified analyses for our data are limited in statistical power, there are indications that risk estimates for Poland are higher than those for other countries (data not shown).
Studies on laryngeal cancer subsites are not available in Central Europe but have been reported for other regions. In a Western European multicenter study, Tuyns et al. (10) reported higher odds ratios for supraglottic cancers in the higher categories of smoking and for glottic/subglottic cancers in the higher categories of alcohol drinking. Muscat and Wynder (11) reported that odds ratios for both cigarette smoking and alcohol drinking were higher for supraglottic cancer compared with glottic cancer, in a study in the United States. Menvielle et al. (12) reported that subsite risks did not differ by tobacco use but were higher for supraglottic cancers for alcohol consumption in France. In a study from Uruguay, De Stefani et al. (13) reported higher odds ratios of ever smoking and ever drinking for supraglottic cancer than for glottic cancer. Our results are consistent with those of previous studies that generally report higher point estimates for the supraglottic region for tobacco and alcohol compared with the subglottic or glottic regions. Statistically significant differences in the laryngeal subsites have been difficult to demonstrate.
An important finding from our study was the moderate effect observed for alcohol consumption and a greater than multiplicative interaction between alcohol and tobacco use. It is possible that we did not detect an increase in risk due to alcohol among nonsmokers, because of the small number of nonsmoking cases in our study, although it is also plausible that alcohol consumption on its own is not an important risk factor for laryngeal cancer. Our finding of a synergistic interaction between tobacco and alcohol on the risk of laryngeal cancer is consistent with findings from previous studies (4) and supports the hypothesis that alcohol acts primarily as a solvent for tobacco carcinogens.
Our study is subject to several limitations, such as the use of patients as controls. Because of the use of hospital controls, our control group may not reflect the general base population from which the cancer cases arose, even though patients were diagnosed with diseases unrelated to tobacco and alcohol. On the other hand, using patients as controls may be beneficial in limiting potential recall bias, because they may put as much effort into recalling their lifestyle habits as cancer cases do. Another potential limitation is that bias may be introduced, because the interviewers knew the case or control status of the subject they were interviewing. This possible bias is unlikely to have been important, given that all interviewers underwent extensive training in order to collect case and control information in an identical fashion. Furthermore, they were unaware of any particular hypothesis being tested.
In summary, tobacco use was a strong risk factor for supraglottic and glottic cancers in Central and Eastern Europe, while alcohol consumption in combination with tobacco may also confer important risks. Approximately 87 percent of laryngeal cancer cases in Central and Eastern Europe are attributable to tobacco use, 75 percent among current smokers and 12 percent among former smokers. Preventive efforts to encourage current smokers to quit tobacco use are likely to be the most effective way to reduce the incidence of laryngeal cancer in this region.
|
ACKNOWLEDGMENTS |
|---|
M. H. worked on this study partly under the tenure of a Special Training Award from the International Agency for Research on Cancer. This work was supported by a grant from the European Commission's INCO-COPERNICUS Program (contract IC15-CT98-0332).
Conflict of interest: none declared.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
- Ferlay J, Bray F, Pisani P, et al. (2001) GLOBOCAN 2000: cancer incidence, mortality and prevalence worldwide, version 1.0 (IARC Press, Lyon, France) (IARC CancerBase no. 5).
- Bray I, Brennan P, Boffetta P. (2000) Projections of alcohol- and tobacco-related cancer mortality in Central Europe. Int J Cancer 87:122–8.[CrossRef][ISI][Medline]
- Alcohol drinking. (1988) IARC Working Group, Lyon, 13–20 October 1987. IARC Monogr Eval Carcinog Risks Hum 44:1–378.[Medline]
- Tobacco smoke and involuntary smoking. (2004) IARC Monogr Eval Carcinog Risks Hum 83:1–1438.[Medline]
- Scelo G, Constantinescu V, Csiki I, et al. (2004) Occupational exposure to vinyl chloride, acrylonitrile and styrene and lung cancer risk (Europe). Cancer Causes Control 15:445–52.[CrossRef][ISI][Medline]
- Mannetje A, Fevotte J, Fletcher T, et al. (2003) Assessing exposure misclassification by expert assessment in multicenter occupational studies. Epidemiology 14:585–92.[CrossRef][ISI][Medline]
- Greenland S and Drescher K. (1993) Maximum likelihood estimation of the attributable fraction from logistic models. Biometrics 49:865–72.[CrossRef][ISI][Medline]
- Vineis P, Alavanja M, Buffler P, et al. (2004) Tobacco and cancer: recent epidemiological evidence. J Natl Cancer Inst 96:99–106.
[Free Full Text] - Zatonski W, Becher H, Lissowska J, et al. (1991) Tobacco, alcohol, and diet in the etiology of laryngeal cancer: a population-based case-control study. Cancer Causes Control 2:3–10.[Medline]
- Tuyns AJ, Esteve J, Raymond L, et al. (1988) Cancer of the larynx/hypopharynx, tobacco and alcohol: IARC international case-control study in Turin and Varese (Italy), Zaragoza and Navarra (Spain), Geneva (Switzerland) and Calvados (France). Int J Cancer 41:483–91.[ISI][Medline]
- Muscat JE and Wynder EL. (1992) Tobacco, alcohol, asbestos, and occupational risk factors for laryngeal cancer. Cancer 69:2244–51.[CrossRef][ISI][Medline]
- Menvielle G, Luce D, Goldberg P, et al. (2004) Smoking, alcohol drinking and cancer risk for various sites of the larynx and hypopharynx. A case-control study in France. Eur J Cancer Prev 13:165–72.[CrossRef][ISI][Medline]
- De Stefani E, Boffetta P, Deneo-Pellegrini H, et al. (2004) Supraglottic and glottic carcinomas: epidemiologically distinct entities? Int J Cancer 112:1065–71.[CrossRef][ISI][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||