Search tips
Search criteria 


Logo of bmjLink to Publisher's site
BMJ. 1999 May 8; 318(7193): 1289.
PMCID: PMC1115672

Alcohol intake and cancer of the upper digestive tract

Pattern of risk in Italy is different from that in Denmark
Carlo La Vecchia, Associate professor of epidemiology
Istituto di Statistica Medica e Biometria, Università di Milano, 20133 Milan, Italy
Silvia Franceschi, Consultant epidemiologist
Adriano Favero, Staff scientist
Renato Talamini, Registrar
Centro Regionale Oncologico, 33081 Aviano PN, Italy
Eva Negri, Head, unit of epidemiological methods

Editor—A Danish study of 156 subjects with upper digestive tract cancers showed that wine drinkers may be at a lower risk than drinkers with a similar intake of beer or spirits.1 Wine is the most common alcoholic beverage in Italy, accounting for over 80% of alcohol intake.2 We investigated the separate and combined effect of wine drinking in a large dataset.3

Data were collected between 1984 and 1993 in Greater Milan and the province of Pordenone, Northern Italy, on 547 histologically confirmed incident cases of cancer of the oral cavity and pharynx (466 men, 81 women; age 22-74 (median 57)) and 412 cases of cancer of the oesophagus (345 men, 67 women; age 26-74 (median 60)). Controls were 2375 subjects (1797 men, 578 women; age 25-74 (median 56)) admitted to the same network of hospitals with acute non-neoplastic conditions unrelated to alcohol or tobacco consumption. Of these, 570 were admitted for trauma, 593 for non-traumatic orthopaedic conditions, 475 for acute surgical diseases, 428 for eye diseases, and 309 for miscellaneous other illnesses. Trained interviewers questioned cases and controls, using a structured questionnaire; questions on the days per week that each type of alcoholic beverage (wine, beer, and spirits) was consumed and the average number of drinks per day were included.

The table gives the distribution of cases of oral and pharyngeal and oesophageal cancer and of the comparison group according to consumption of various alcoholic beverages. When non-drinkers and moderate drinkers (<3 drinks a day) were used as the comparison group, the odds ratios of oral and pharyngeal cancer were 4.95 for very heavy ([gt-or-equal, slanted]10 drinks per day) wine drinkers and 4.13 for very heavy wine, beer, and spirit drinkers. Corresponding findings for oesophageal cancer were 7.53 and 5.52. The odds ratios for beer or spirit drinkers only were 0.9 (95% confidence interval 0.3 to 2.4) for cancers of the oral cavity (7 cases, 47 controls) and 1.5 (0.6 to 3.9) for oesophageal cancer (7 cases, 47 controls). These results were consistent across separate diagnostic conditions in controls.

This pattern of risk according to type of alcoholic beverage is different from that reported in the Danish study.1 Taken together, these data lead us to conclude that ethanol is the main component of alcoholic beverages that determines the risk of cancer and that the most frequently consumed beverage in each area tends to be the one with the highest risk.4,5 Wine consumption is clearly strongly related to cancers of the upper digestive tract in individuals and populations with heavy consumption.

Distribution of 547 cases of oral and pharyngeal cancer, 412 cases of oesophageal cancer, and 2375 controls, and corresponding multivariate odds ratios* and 95% confidence intervals, according to total number of drinks per week of wine only and ...


1. Grønbæk M, Becker U, Johansen D, Tønnesen H, Jensen G, Sørensen TI. A population based cohort study of the association between alcohol intake and cancer of the upper digestive tract. BMJ. 1998;317:844–848. . (26 September.) [PMC free article] [PubMed]
2. La Vecchia C. Alcohol in the Mediterranean diet. Assessing risks and benefits. Eur J Cancer Prev. 1995;4:3–5. [PubMed]
3. Barra S, Franceschi S, Negri E, Talamini R, La Vecchia C. Type of alcoholic beverage and cancer of the oral cavity, pharynx and oesophagus in an Italian area with high wine consumption. Int J Cancer. 1990;46:1017–1020. [PubMed]
4. Doll R, Forman D, La Vecchia C, Woutersen R. Alcoholic beverages and cancers of the digestive tract and larynx. In: Verschuren PM, editor. Health issues related to alcohol consumption. Washington: International Life Sciences Institute; 1993. pp. 126–166.
5. International Agency for Research on Cancer. Alcohol drinking. Lyons: IARC; 1988. (IARC monographs on the evaluation of carcinogenic risks in humans volume 44.)
BMJ. 1999 May 8; 318(7193): 1289.

Confounding in such studies is important

K K Cheng, Professor of epidemiology
C Cummins, Lecturer
R Maric, Research fellow

Editor—Grønbæk et al concluded that moderate consumption of wine does not increase the risk of cancer of the upper digestive tract whereas a similar intake of beer or spirits does.1-1 This is a potentially important finding, at least for those of us who mostly drink wine. But several issues concerning confounding warrant some discussion.

Firstly, there seemed to be fewer heavy drinkers (>21 drinks/week) in the group in which wine constituted >30% of total alcohol intake than in the group in whom it constituted 0% of the intake (10.2% v 31.2%, from data in table 2). It has been documented that there is probably little increase in risk of oesophageal cancer below an intake of 200 g/week (16.7 drinks from Grønbæk et al’s definition).1-2 It would have been useful if the authors had subdivided the >21 drinks/week category to examine the risks associated with higher intake so that residual confounding could be excluded, although such an attempt may be hampered by small numbers.

Another source of potential residual confounding is the amount of tobacco smoked. Again, to adjust properly for any confounding effects arising from smoking one would like to see finer divisions of intake than those reported in the paper (1-19 g/day and 20 g/day).

Finally, the authors mentioned the absence of data on diet. It has been shown consistently that eating fresh fruit and vegetables is strongly protective against oesophageal cancer.1-3 In the West Midlands a recent survey on lifestyles showed that wine drinkers tended to have a better diet than those who drank beer or spirits (unpublished data). If this was also the case in Denmark then it could explain at least part of the difference observed. The authors’ case would be stronger if they can show that such associations do not exist in Denmark.

On the whole, we would agree with the editorial accompanying the paper that there is “still no clear evidence to link specific beverages to specific cancers.”1-4


1-1. Grønbæk M, Becker U, Johansen D, Tønnesen H, Jensen G, Sørensen TIA. Population based cohort study of the association between alcohol intake and cancer of the upper digestive tract. BMJ. 1998;317:844–847. . (26 September.) [PMC free article] [PubMed]
1-2. Cheng KK, Duffy SW, Day NE, Lam TH, Chung SF, Badrinath P. Stopping drinking and risk of oesophageal cancer. BMJ. 1995;310:1094–1097. [PMC free article] [PubMed]
1-3. Cheng KK, Day NE. Nutrition and esophageal cancer. Cancer Causes Control. 1996;7:33–40. [PubMed]
1-4. Sabroe S. Alcohol and cancer. BMJ. 1998;317:827. . (26 September.) [PMC free article] [PubMed]
BMJ. 1999 May 8; 318(7193): 1289.

Such studies should be done in non-smokers

Vaughan Reed, Freelance statistician

Editor—Grønbæk et al’s study is a further example of a large epidemiological study in which smokers have been included even though smoking was likely to be a risk factor for the disease under study.2-1 Statistical manipulation was then used to correct for the effects of smoking. But how reliable are these corrections, and how much faith should be put in the answers obtained with regard to the specific question of alcohol consumption causing cancer? Given that corrections for relatively straightforward data are often only approximate, the development of cancer is unlikely to be accurately corrected for, particularly when smoking habit is only broadly classified.

The term “correction for smoking” is ambiguous. In some studies authors have clearly analysed out the effects of smoking and extrapolated back, in order to predict values for non-smokers. The present authors seem to have corrected to a standard value for smoking, probably the median of their categories. If this interpretation is correct then their graphs and tables show the morbidity standardised for a population who had formerly smoked for at least five years.

The authors suggest that smoking was not the main influence on morbidity since the level of cancer in the heavy smokers (risk 7.1) was lower than that in the heavy drinkers (risk 11.7). However, their figure 1 indicates that this risk in heavy smokers was in fact higher than that for all but one of the alcohol consumption categories, which implies that a large proportion of the morbidity may still have occurred in the smokers. They also point out that heavy smoking and heavy drinking tended to go together, which makes accurate correction even harder.

Many non-smokers are treated with drugs to reduce their cholesterol concentration or blood pressure, largely on the evidence of studies that have included smokers and in which smokers have contributed most of the morbidity. The justification for treating non-smokers is generally based on an extrapolation that finds a small residue of morbidity even after smoking has been corrected for. If those corrections overestimate the true risk to non-smokers then these patients might be taking long term medication that may not benefit them since they were not at risk in the first place. We know that smoking is a risk factor for most respiratory and cardiovascular disease as well as for most cancers and some gastrointestinal disorders. Should we not perform epidemiological studies specifically on non-smokers or, at the very least, request authors to provide summary tables of their main findings specifically for the non-smoking subset of their population?


2-1. Grønbæk M, Becker U, Johansen D, Tønnesen H, Jensen G, Sørensen TIA. Population based cohort study of the association between alcohol intake and cancer of the upper digestive tract. BMJ. 1998;317:844–847. . (26 September.) [PMC free article] [PubMed]
BMJ. 1999 May 8; 318(7193): 1289.

Treating upper digestive tract cancers as a single entity may be misleading

Rebecca Fitzgerald, Lecturer in gastroenterology
St Bartholomew’s and the Royal London School of Medicine and Dentistry, London E1 2AD ; rcfitzgerald/at/>
Christine Caygill, Epidemiologist

Editor—Grønbæk et al reported that the type of alcohol consumed may be important in determining the risk of cancer of the upper digestive tract in their population based study in Copenhagen.3-1 In this large epidemiological study details about the alcohol intake and the particular type of alcoholic beverage were obtained by questionnaire. Patients with cancer of the upper digestive tract were then identified from the Danish cancer registry. This registry subdivides oral cavity, pharyngeal, and oesophageal cancers, but the authors make no comment about whether their results applied to all cancers of the upper digestive tract uniformly or not. This information may be important considering the different pathogenesis of these cancers.

Oesophageal cancers, in particular, are made up of two main groups with very different pathogenesis. Squamous carcinoma is declining in incidence in the Western world and is strongly associated with alcohol and tobacco intake, whereas adenocarcinoma (which is associated with the premalignant condition Barrett’s oesophagus) is rapidly increasing in incidence and does not appear to be so strongly associated with smoking and alcohol, although further information is needed.3-2 We appreciate that in the Danish study it would have been difficult for the authors to obtain information on the types of oesophageal cancer, since the cancer registry does not separate the histopathological subtypes. In a Medline search (1995-8) on the relation between oesophageal cancer and alcohol only five of 12 papers specified which histopathological subtype they were referring to.

It is time that cancer registries classified oeosphageal adenocarcinoma separately from squamous carcinoma so that we can gain meaningful information about the epidemiology of these important conditions.


3-1. Grønbæk M, Becker U, Johansen D, Tønnesen H, Jensen G, Sørensen TIA. Population based cohort study of the association between alcohol intake and cancer of the upper digestive tract. BMJ. 1998;317:844–847. . (26 September.) [PMC free article] [PubMed]
3-2. Vaughan TL, Davis S, Kristal A, Thomas DB. Obesity, alcohol and tobacco risk for cancers of the esophagus and gastric cardia: adenocarcinoma vs squamous cell carcinoma. Cancer Epidemiol, Biomarkers and Prev. 1995;4:85–92. [PubMed]
BMJ. 1999 May 8; 318(7193): 1289.

Authors’ reply

Morten Grønbæk, Senior research fellow

Editor—The large number of cases in La Vecchia et al’s case-control study enables them to present a detailed analysis of the effect of wine only and wine together with other types of drinks. Their study shows what ours cannot—namely, that even with wine there is an increased risk of the cancers. As Cheng et al point out, our study was hampered by a smaller number of cases in the heavy drinking categories. Unfortunately, La Vecchia et al’s study does not provide an estimate for non-drinkers and does not distinguish the effect of drinking wine from the effect of drinking beer and spirits.

Both Cheng et al and Reed are concerned about the possible confounding effect of smoking, which we controlled for quite thoroughly. In one analysis (reported in the paper) we controlled for smoking by taking into account differences in the prevalence of smoking among individuals at different levels of wine percentages and of total alcohol intake in the Cox regression models. In another model (not reported) we took into account the number of pack-years and inhalation; this yielded the same results as those in the paper. The number of cases did not allow stratified analyses of smokers and non-smokers. However, the aim of the study was not to assess smoking or alcohol intake as risk factors for cancer of the upper digestive tract (this has been extensively studied) but to search for potential differences in the effects of the three types of alcohol. Unless the effect of smoking and type of alcoholic beverages interact, the need to estimate risk among smokers and non-smokers is less prominent.

As Cheng et al point out, a few studies have shown a protective effect of fruit and vegetable intake on oesophageal cancer. To be a confounder in the present analysis, however, the effects of intake of fruit and vegetables and its association with wine intake would have to be unexpectedly strong to explain our findings.

If we had had a larger number of cases we would have been able to distinguish between these types of cancer and even the histopathological subtypes, as suggested by Fitzgerald and Caygill. If we misclassified some of the cases we would expect this to have reduced the differences in the effects of beer, wine, and spirits.

Articles from BMJ : British Medical Journal are provided here courtesy of BMJ Group