Carcinogenetic impact of alcohol intake on squamous cell carcinoma risk of the oesophagus in relation to tobacco smoking

Abstract

Consumption of alcohol and tobacco, separately or jointly, can increase the risk of oesophageal squamous cell carcinoma (OSCC). It is unclear whether the amount of alcohol consumption by individual drinkers affects the joint carcinogenetic action of both agents. To demonstrate how the intensity of alcohol intake determines the risk of OSCC in relation to tobacco smoking, we conducted a multicentre case-control study. A total of 652 patients with pathology-proven OSCC, as well as 1127 gender, age, and study hospital matched controls were recruited. To identify a possible curvature in the continuous relationship between exposure and risk, we applied the generalised additive models to the collected data. Both non-drinkers who smoked tobacco and non-smokers who drank heavy alcohol (>30 g/day) were observed to have elevated cancer risks. A smoking habit-specific, non-linear increase in oesophageal cancer risk was recognised. Tobacco was found to interact with light-to-moderate alcohol (0.1–30 g/day) to increase the risk of oesophageal cancer in a supra-multiplicative way (Odds ratio (OR) ratio = 5.5–5.7, p < 0.05), whereas with heavy alcohol consumption in a simple multiplicative model (OR ratio = 1.7–2.3, p > 0.05). Weekly intake frequency had the strongest influence on the risk of neoplasm development. Alcohol consumption was responsible, respectively, for 18% and 77% of nonsmoking and smoking OSCC cases in this population. In conclusion, both light-to-moderate and heavy alcohol intake interact separately with tobacco in differently synergistic processes that can determine the development of this type of cancer.

Introduction

Heavy alcohol consumption has been demonstrated to be a major health hazard.¹ However, the consumption of light-to-moderate amounts of alcohol may not be detrimental, and may even be beneficial to the prevention of coronary heart disease and stroke.2, 3 Still, the effects of low-to-moderate alcohol consumption on other diseases, such as cancer, have not been well established.

Consumption of alcohol and tobacco, separately and together, are two of the most important determinants of carcinoma of the oesophagus.4, 5 Most studies have recognised that the combined effect of these two agents is more than additive, and this is often illustrated adequately by a simple multiplicative model.⁶ However, evidence of supra-multiplicative interaction has been discovered in some earlier reports.7, 8 While ethanol itself has not been found to cause cancer in animal experiments,⁴ it is, in general, considered either as a solvent for other active carcinogens, or as an enhancing factor.⁴ It is uncertain whether the level of ethanol to which a subject is exposed might affect such joint carcinogenetic action.

Among men in Taiwan, the age-standardised incidence rate has been reported as 8.7 per 100,000 by the population-based cancer registries in 2002;⁹ this is compatible to numbers found in intermediate and high-risk areas of central Europe and South America.¹⁰ In regions such as northern France11, 12 and northern Italy,13, 14 where heavy alcohol intake is common, it is difficult to evaluate the effect of light-to-moderate alcohol consumption on oesophageal cancer. In Taiwan, where heavy alcohol consumption is less common, lifestyle and dietary factors may differ from those of the western populations.¹⁵ Therefore, those special epidemiological characteristics warrant further research regarding the effects of alcohol consumption upon the genesis of oesophageal carcinoma within the Taiwanese population.

In the evaluation of dose-response relationship between alcohol intake and oesophageal cancer risk, categorical analysis, which assumes the risk is constant inside each category, was widely employed in earlier studies that modelled such associations.¹⁶ Log-linear dependence, i.e. a proportional increase in log-risk in these models, was presumed. However, as with most biological measures, a non-linear effect of alcohol on cancer risk could be expected.¹⁷ Further, data categorisation might induce some potential biases in the step function analysis.18, 19 In the process of looking for a possible curvature in the continuous relationship between exposure and risk, we employed the technique of generalised additive models (GAM), which assumes that the mean of the dependent variable depends on an additive predictor through a nonlinear link function.²⁰

The objectives of this study are 2-fold: first, to explore the carcinogenetic impact of different aspects of alcohol intake on the development of oesophageal cancer in relation to tobacco consumption, and second, to employ a more flexible approach (GAM) to the study of the dose-dependent relationship between the intensity of alcohol consumption and the risk of this neoplasm.