Alcohol drinking is an important determinant for many causes of death, including cancer.1, 2 In many populations, a strong association is observed between socioeconomic position (SEP) and alcohol related mortality with higher mortality among subjects with a low SEP.3, 4, 5 With regards to mortality from specific cancers related to alcohol use (liver, larynx, oral cavity, pharynx, oesophagus), however, variations in the level of socioeconomic inequalities among men are found between European populations. Large inequalities are found in Spain and Italy and, especially, in France.6, 7, 8, 9, 10 On the contrary, some studies have suggested small socioeconomic inequalities in the Nordic countries and Switzerland.10, 11, 12
Nevertheless, the literature is rather scarce and a European overview of differences in socioeconomic inequalities in alcohol related cancers is currently lacking. It would be of interest to document the true extent of the problem within Europe. Contrary to smoking,13 the role of alcohol in socioeconomic inequalities in cancer mortality has not yet been evaluated but may also be important. In addition, a comparison between European populations would show whether different patterns in socioeconomic inequalities in alcohol-related cancers are found within Europe and whether these patterns could be related to different drinking cultures. Differences in national levels of alcohol related cancers mortality rates are found between Western European countries, with substantially higher rates in Spain and Italy, and especially in France.14 In addition, different drinking cultures are observed in Western Europe between countries but also within some countries. Daily wine consumption especially during meal is more common in countries like Spain, Portugal, Italy or France or in the Latin part of Switzerland whereas binge drinking and beer consumption is more widespread in the UK, the Nordic countries and the German part of Switzerland.15, 16, 17
The aim of this study was to investigate differences in socioeconomic inequalities in alcohol-related cancers mortality between Western European populations. Our dataset included longitudinal studies from 13 populations from South to North of Western Europe with information on causes of death. We included populations with contrasted situations with regards to overall levels of alcohol related cancers mortality rates and drinking cultures.
We will first investigate socioeconomic inequalities among men in alcohol related cancers mortality, and thereby distinguish liver and upper aerodigestive tract (UADT) cancers. We will then focus on the contribution of alcohol related cancers to socioeconomic inequalities in mortality from all cancers types together. As UADT cancers are also smoking related, we will finally study lung cancer as an indicator of the smoking situation in each population and evaluate to which degree the international patterns of inequalities in UADT cancers are correlated with those for lung cancer.
Material and methods
Longitudinal data from 13 European populations were used, including Madrid, the Basque region (Spanish part), Barcelona, Turin, Switzerland (Latin and German part), France, Belgium (Brussels, Walloon and Flemish part), Sweden, Norway and Finland. Most datasets covered the entire national population, except for France (a representative sample of 1% of the population), Madridand Basque region (regions), Barcelona and Turin (urban areas). In Belgium and Switzerland, we distinguished regions with differences in alcohol consumption and drinking patterns that could induce differences in socioeconomic inequalities in alcohol related cancers. Men were selected at the time of the population census and followed up during the 1990s (Table I).
Table I. Descriptive Information on the Data Sources
Number of person years at risk
Educational level (%)
Date of census
End of follow-up
Lower secondary or less
Analyses included men aged 30–74 at the census. The follow-up period was shorter for Belgium, Madrid and the Basque region. To have results on comparable ages in terms of observed ages at death, analyses for these 3 populations were conducted on slightly older age groups at baseline (35–79 for Madrid and 30–79 for Belgium and the Basque region).
The linkage between census data and mortality registries was achieved for more than 96% of all deceased persons in all populations except for Madrid (70%), the Basque region (93%) and Barcelona (94.5%). In these latter populations, however, no variation in this percentage was found according to age, sex or SEP (In the Basque region this check could unfortunately not be done for socioeconomic status). To avoid an underestimation of absolute mortality rates in these 3 populations, observed absolute mortality were increased by correction factors (1/0.70, 1/0.93 and 1/0.945, respectively).
The SEP was measured with education declared at the time of the population census. This variable was categorized into 3 classes that corresponded to the ISCED (International Standard Classification of Education) classification: 0–2 (lower secondary education or less), 3–4 (upper secondary education), 5–6 (postsecondary education). The percentage of missing values for education was of 17% in Brussels, 5% in the Walloon and Flemish parts of Belgium and less than 3% for all other populations. These subjects were excluded from the analysis.
The cause of death was obtained by linkage with death registries. Analyses were conducted for all cancer mortality (ICD 9: 140–249), for lung cancer (ICD 9: 162–163, 165), and for alcohol related cancers: UADT [that groups oral cavity, pharynx, esophagus and larynx (ICD 9: 140–150, 161)] and liver (ICD 9: 155). UADT and liver cancer were selected for analyses because they are strongly associated with alcohol consumption1, 18 and because they presented a substantial population attributable fraction (PAF) for alcohol (20–40% for UADT and 32% for liver1, 19). Lung cancer was selected as an indicator for the cumulative exposure of the population to smoking. This approach is considered to be acceptable, although lung cancer mortality is only an approximate indicator.20
The magnitude of socioeconomic inequalities in mortality was estimated in both absolute and relative terms. To estimate relative inequalities, we computed relative indices of inequality (RII), using Poisson regression. The calculation of the RII is based on a ranked variable, which specifies for each educational group the mean proportion of the population with a higher level of education. For instance, the rank of the lowest educational group is calculated as the proportion of the population with middle or high education, plus half of the proportion of the population with a lowest educational level. The RII is then computed by regressing the mortality on this ranked variable. Thus, the RII expresses inequality within the whole socioeconomic continuum. It can be interpreted as the ratio of mortality rates between the 2 extremes of the educational hierarchy. As it takes into account the size and relative position of each educational group, it is well adapted to compare populations with different educational distributions.21, 22
To estimate absolute socioeconomic inequalities we computed absolute rate differences between the lowest and the highest educational level, both for all cancer mortality and for the specific cancer types. Age-standardized mortality rates were computed, using the population of EU-15 plus Norway of 1995 as the standard population. The contribution of these different cancer types to socioeconomic inequalities in all cancer mortality was also calculated by expressing the rate difference for this cancer type as a percentage of the rate difference for all cancer mortality.
The educational distributions highly differed between the populations (Table I). The percentage of subjects with lower secondary education or less was the highest in the 3 Spanish populations and Turin (around 65%) and the lowest in Norway (less than 30%) and Switzerland (around 20%).
For UADT and liver cancers, we observed a regular inverse gradient in cancer mortality, with higher cancer rates for lower educational levels (Fig. 1). Differences were found between populations and the situation is remarkable in France with the highest mortality rate among men with lower secondary education or less and among the lowest mortality rate among men with postsecondary education.
For UADT cancers, the largest RII's (above 3.5) were observed in France, Switzerland (German and Latin part) and Turin (Table II). The RII was lower than 2 but still significant in Belgium. For liver cancer, the largest RII's (above 2.5) were found in Madrid, France and Turin. In contrast, the RII was around 1 and nonsignificant in the Basque region, Belgium and Norway. For lung cancer, the largest RII's (around 3 or above) were observed in Finland, Belgium and the German part of Switzerland. They were lower than 2 but still significant in the Spanish populations and France.
Table II. Mortality Rates (MR) and Relative Indices of Inequality (RII) with Their 95% Confidence Intervals (CI) for all Cancers and by Cancer Type, Per Population
Age-standardized mortality rate using direct standardization, per 100,000 person years.
Absolute mortality rate differences by cancer site are presented in Figure 2. The most striking result is the large range of mortality rates differences found for UADT cancers: from 7 in Sweden to 78 per 100,000 person years in France. It was between 20 and 40/100,000 in the Spanish populations and Switzerland (German part) and 46 in Switzerland (Latin part). The contribution of these cancer sites to socioeconomic inequalities is presented in Table III. The contribution of UADT cancers was the highest in France, Barcelonaand the Basque region (around 25%), followed by the Latin part of Switzerland and Madrid (20%). The contribution of liver cancer was much lower. However, we observed differences between populations with the largest contribution in Madrid (15%) and also a substantial contribution in France, Barcelona and Turin (9%) whereas it was lower than 6% in all other populations. All in all, the contribution of alcohol related cancer to socioeconomic inequalities in cancer mortality was 29–36% in France and the Spanish populations, 17–23% in the Swiss regions and Turin, and 5–15% in Belgium and the Nordic countries.
Table III. Contribution (%) of Different Cancer Sites to Absolute Socioeconomic Inequalities in Cancer Mortality, Per Population
We do not observe a correlation between absolute inequalities for lung and UADT cancers (Fig. 3). Populations with the largest inequalities in lung cancer are not those with the largest inequalities in UADT cancers. Belgium shows large rate difference for lung cancer but small difference for UADT cancers. The rate difference for UADT cancers is similar in Madrid and the German part of Switzerland, whereas the rate difference for lung cancer is 2 times lower in Madrid. France shows particularly high difference for UADT cancers but only a medium rate difference for lung cancer. Also in terms of RII's, there is no correlation between mortality rates differences for lung and UADT cancers (see Table II).
This study focused on differences between Western European populations with regards to socioeconomic inequalities in alcohol related cancer mortality. Large differences were found. Inequalities were largest in Spain, Switzerland and France and smallest in the Nordic countries and Belgium. In France, socioeconomic inequalities were remarkably large for UADT cancers. The contribution of alcohol-related cancers to socioeconomic inequalities in cancer mortality was high in France, Madrid and Barcelona (35%) compared to small (less than 5–15%) in Belgium and the Nordic countries. The lack of correlation between the inequalities found for lung and UADT cancers suggested that, even though smoking is a major risk factor for UADT cancers, large inequalities in UADT cancers were also due to other factors, probably alcohol drinking.
Evaluation of data
There are differences in the follow-up periods. Given the shorter follow-up period in Madrid, the Basque region and Belgium, we changed the age range at baseline for these populations such that studies were similar in terms of average at death. However, subjects may be slightly older or younger in these populations compared to others. This could have resulted in a slight underestimation of relative socioeconomic inequalities and overestimation of absolute inequalities for these populations. Nevertheless, these effects, if any, are likely to be small.
Some differences were found in the populations covered. In France and Switzerland, foreigners were excluded and analyses were thus conducted for more homogeneous populations. Perhaps the exclusion of foreigners has lead to underestimation of inequalities in alcohol-related cancers mortality in these countries. A large part of migrants, at least in France, come from Muslims countries and often do not drink alcohol for religious reasons.23 In France, they generally have low levels of UADT cancers mortality. For liver cancer, on the other hand, mortality rates among migrants are higher than in the native population but the etiology is different (due to Hepatitis B or C infection).24, 25, 26
Differences could occur between populations in the coding of causes of death. Even though data came from countries with reliable cause-of-death registries, national diagnosing practices may differ between countries. International comparisons revealed that more deaths were classified as cancer deaths in France than in other countries, probably leading to an overestimation of the French cancer mortality rates.27, 28 This bias could be a serious issue for absolute measures of inequalities, especially if it occurs more for some cancer sites (for instance UADT in France). With regards to relative measures of inequalities, our results would be biased only if diagnosing practices differ by SEP of the deceased, and if this applies especially to some cancer sites. There is no evidence to support this hypothesis.
In addition, there is a specific problem related to liver cancer mortality rates because of frequent misclassification of metastases as primary cancers. An American study suggested that between 27 and 31% of liver cancer deaths were due to metastases or secondary cancers instead of primary cancers.29 The results relating to liver cancer should therefore be considered cautiously. Unfortunately, no study investigated the potential association between SEP and misclassification as well as possible variations between countries. If the rate of misclassification does not differ by SEP, this problem would impact on absolute inequalities but not on relative inequalities.
Socioeconomic status was measured using information on educational level. We used a common classification for all populations that should avoid problems with the comparability between educational systems of different countries. However, large differences were observed between populations in the educational distributions. Part of these differences may be due to real differences in educational levels. But we cannot rule out the possibility that there are differences in the way in which educational systems are being squeezed into this common classification. However these differences probably have a weak influence on the results found here. We evaluated the sensitivity of the results to alternative educational classifications. In one type of analyses, for example, we used a classification into 4 educational levels by distinguishing between men who completed lower secondary education from men with primary education only. We also considered another classification in 3 educational levels to get population distributions that were as similar as possible between populations. The results obtained with these alternative classifications were quite similar to those presented here.
Several European countries were not included in this analysis. We did not include any country from Eastern Europe because of lack of longitudinal mortality studies. We also did not include the UK since British data were not accessible for small causes of death because of confidentiality rules. In the UK, a low contribution of non-lung cancers to socioeconomic inequalities in all cancer mortality was found in the 1980s30 whereas this contribution was comparable to that of lung cancer in another study conducted in the 1990s.31
Possible explanations of the results
Socioeconomic inequalities in the distribution of risk factors may largely explain the results. Smoking is a major risk factor for both lung and UADT cancers; the PAF for smoking for UADT cancers is indeed around 70%. Therefore, smoking may potentially explain a large part of the observed inequalities in UADT mortality. Nevertheless, smoking alone cannot fully explain the international patterns in inequalities. The differences in inequalities in UADT cancer between countries with comparable inequalities in lung cancer mortality, especially between northern and southern countries, point to the effect of other factors. Given the high PAF found for alcohol (between 20 and 40%)1, 19 for UADT cancers, alcohol consumption is certainly one of those factors.
Consequently, variations in drinking patterns between European populations may partly explain our results. It is unlikely that the type of alcohol accounts for the differences observed as the type of alcohol consumed does not seem to have an effect on risk of UADT cancers.32, 33 Differences in socioeconomic inequalities in the total amount of alcohol consumed may be the critical factor. In general, excessive alcohol consumption is found to be higher in men with low SEP, although results differ according to the country. No inequalities in high alcohol consumption are observed in Belgium34 and inequalities are consistently reported for France.35, 36 A European study suggested that France was the country with the largest inequalities, but only for excessive consumption (more than 6 drinks per day).34 Some studies do not report inequalities in Northern Italy37 or Barcelona,38 whereas a European study observed inequalities in Spain, but only for excessive consumption.34 In Sweden, higher alcohol consumption was found among non-manual workers in the 1970s, but an equalization of the social differences in heavy drinking and a tendency to reversal were observed in later years.39 In Norway, higher alcohol consumption was observed in the upper education and income groups.40 Thus, even though the literature is not totally consistent, it is globally in accordance with our findings.
With regards to absolute inequalities, the absolute level of consumption has also to be taken into account. It is higher in France, Spain, Switzerland, Italy followed by Belgium and lower in Finland and especially low in Sweden and Norway.41 France thus presents a combination of both high level of alcohol consumption and relatively large inequalities in this consumption, followed by Spain. This may explain the largest absolute inequalities found in these populations.
We distinguished in Belgium and Switzerland different regions that could have relevant differences in drinking cultures. Whereas alcohol related cancer mortality rates gave a consistent “cultural” pattern with higher rates in the Walloon part of Belgium and especially in the Latin part of Switzerland, we observed in these regions only a slightly higher contribution of alcohol related cancers to inequalities in total cancer mortality. Few studies have been conducted on drinking pattern by linguistic region. They found no clear variations in Belgium42 but higher daily and wine consumption in the Latin part of Switzerland compared to the German part.17 Our results for Belgium and Switzerland are thus globally consistent with these studies and with our results found in the bordering countries. However, we could have expected more pronounced differences between linguistic regions. It seems that between regions within the same country, the pattern of socioeconomic inequalities in cancer mortality only slightly differed around a global national pattern. These results suggest that national factors, such as common national histories, socioeconomic policies and health care systems, predominate over regional factors in determining socioeconomic inequalities in cancer mortality.
The situation in France is remarkable because the large socioeconomic inequalities in alcohol related cancers, and especially in UADT cancers. The situation may not be homogeneous within France. Larger inequalities in all-cause mortality are found in French regions with a higher alcohol consumption, in particular in the North.43, 44 This result suggests that there could be large regional disparities in France in inequalities in alcohol-related mortality in general, and in alcohol-related cancer mortality in particular. The small size of our French dataset and resulting the lack of statistical power however hampered a regional analyses for France.
An important and consistent result in our study is that we do not observe large socioeconomic inequalities in alcohol-related cancer mortality in Northern Europe (Belgium and the Nordic countries). Some studies found a strong impact of alcohol drinking on health inequalities in the Nordic countries, but mainly through violent deaths.4, 5 Interestingly, it is also in those countries that binge drinking is more widespread, whereas Spain and France are characterized by higher levels of daily alcohol consumption. These results suggest that binge drinking is mainly associated with inequalities in violent deaths whereas high levels of daily consumption influences inequalities in mortality in part through specific cancers.
Other risk factors than alcohol drinking and smoking may also partly explain our results. Liver cancer is related to infection from Hepatitis B or C, but mainly in countries with high liver cancer incidence, which is not the case of Western Europe.45 Diet46 and occupational exposures47, 48 could also contribute to inequalities in mortality from UADT cancers, but their impact is likely to be weaker than that of alcohol.
Socioeconomic inequalities in cancer survival may partly explain socioeconomic differences in cancer mortality. Survival inequalities may be more important for cancers with a relatively good prognosis compared to cancer with very low survival rates.49, 50, 51 Thus, socioeconomic inequalities in cancer survival might be more important for UADT cancers, as these have a better prognosis than liver or lung.52, 53, 54, 55 Unfortunately, no comparative study is available on differences between European populations in socioeconomic inequalities in cancer survival.
Inequalities in alcohol use has an impact on health inequalities in Europe. This has been shown by studies that found an impact of heavy drinking on socioeconomic inequalities in Northern Europe through poisoning, accidents and suicides. Our study showed that high alcohol consumption also impacts on health inequalities through cancer, but more so in Southern Europe (such as in Spain, France and Switzerland) than in Northern Europe. Thus, while heavy drinking is an important contributor to socioeconomic inequalities in mortality, there are large differences between populations in the relevant consumption patterns and associated causes of death.
Mrs. G. Menvielle received a funding from the Fondation pour la Recherche Médicale for this analysis. The Swiss data are from the Swiss Federal Statistical Office. The construction of the Swiss National Cohort has been supported by the Swiss University Conference (Network Public Health, Swiss School of Public Health).