Odds ratios and 95% confidence intervals were estimated using conditional logistic regression. The exposed subjects were categorized in three groups based on tertiles of the distributions of exposure among controls. Variables included in the adjusted analyses were total pack-years of smoking (5 categories), total coffee consumption (4 categories) and occupational skill and status (4 categories): All jobs in the occupational histories were divided according to major ISCO88 groups (first digit), 1e3: high-skilled white-collar worker, 4e5: low-skilled white-collar worker, 6e7: high-skilled blue-collar worker and 8e9: low-skilled blue-collar worker.