Joint and independent effects of alcohol drinking and tobacco smoking on oral cancer: a large case-control study.

Abstract

Alcohol drinking and tobacco smoking are assumed to have significant independent and joint effects on oral cancer (OC) development. This assumption is based on consistent reports from observational studies, which, however, overestimated the independent effects of smoking and drinking, because they did not account for the interaction effect in multivariable analyses. This case-control study sought to investigate the independent and the joint effects of smoking and drinking on OC in a homogeneous sample of adults. Case patients (N = 1,144) were affected by invasive oral/oropharyngeal squamous cell carcinoma confirmed histologically, diagnosed between 1998 and 2008 in four hospitals of São Paulo (Brazil). Control patients (N = 1,661) were not affected by drinking-, smoking-associated diseases, cancers, upper aero-digestive tract diseases. Cumulative tobacco and alcohol consumptions were assessed anamnestically. Patients were categorized into never/ever users and never/level-1/level-2 users, according to the median consumption level in controls. The effects of smoking and drinking on OC adjusted for age, gender, schooling level were assessed using logistic regression analysis; Model-1 did not account for the smoking-drinking interaction; Model-2 accounted for this interaction and included the resultant interaction terms. The models were compared using the likelihood ratio test. According to Model-1, the adjusted odds ratios (ORs) for smoking, drinking, smoking-drinking were 3.50 (95% confidence interval -95CI, 2.76-4.44), 3.60 (95CI, 2.86-4.53), 12.60 (95CI, 7.89-20.13), respectively. According to Model-2 these figures were 1.41 (95CI, 1.02-1.96), 0.78 (95CI, 0.48-1.27), 8.16 (95CI, 2.09-31.78). Analogous results were obtained using three levels of exposure to smoking and drinking. Model-2 showed statistically significant better goodness-of-fit statistics than Model-1. Drinking was not independently associated with OC, while the independent effect of smoking was lower than expected, suggesting that observational studies should be revised adequately accounting for the smoking-drinking interaction. OC control policies should focus on addictive behaviours rather than on single lifestyle risk factors.