A case–control study on OCP cancer was conducted between 1997 and 2009 in the greater Milan area, Italy and in the canton of Vaud, Switzerland. Cases were 768 patients (593 men and 175 women) under age 79 years (median 58 years, range: 22–79 years) with incident, histologically confirmed squamous cell cancers of OCP (excluding cancers of the lip, salivary glands and nasopharynx), admitted to major teaching or general hospitals in the areas under investigation. Controls were 2078 subjects (1368 men and 710 women, median age 59 years, range: 19–79) with no previous history of cancer, admitted to the same hospitals for acute, nonneoplastic conditions unrelated to tobacco smoking, alcohol drinking or long-term dietary modifications. Among controls, 19% were admitted for traumas, 21% for other orthopaedic conditions, 51% for acute surgical conditions and 9% for other miscellaneous conditions. In Italy, less than 5% of the cases and controls approached refused to participate in the study; in Switzerland, the proportion of refunds was about 15%.
Trained personnel interviewed both cases and controls during their hospital stay using a structured questionnaire, including information on sociodemographic characteristics, anthropometric measures and selected lifestyle habits (including tobacco smoking and alcohol drinking). Subjects' dietary habits during the 2 years before cancer diagnosis or hospitalisation (for controls) were assessed through a valid (Decarli et al, 1996 (link)) and reproducible (Franceschi et al, 1993 (link), 1995 (link)) food frequency questionnaire (FFQ), including information on weekly consumption of 78 foods, recipes and beverages. For a few vegetables and fruits, seasonal consumptions and corresponding durations were elicited. Food items were combined into 18 food groups: milk and yoghurt; cereals; soups; eggs; poultry; red meat; processed meat; fish; cheese; raw vegetables; cooked vegetables; all vegetables; potatoes; citrus fruits; other fruits; all fruits; desserts; and sugars. To estimate the daily intake of nutrients and total energy, we used an Italian food composition database (Gnagnarella et al, 2004 ). To evaluate the role of macronutrients independently from total energy intake, we derived energy-adjusted nutrients according to the residual method (Willett and Stampfer, 1986 (link)).
We categorised food groups, micronutrients and energy-adjusted macronutrients into quintiles (or quartiles/tertiles for a few foods with low frequency of consumption), according to the distribution among the control population. The ORs and corresponding confidence intervals for quantiles of intakes were estimated using multiple logistic regression models, including terms for age (5-year groups), sex, centre (Italy, Switzerland), education (<7, 7 to <12, ⩾12 years), year of interview (continuous), body mass index (BMI, <20, 20 to <25, 25 to <30 kg m−2), tobacco smoking (never smoker, ex-smoker, current smoker of: <15, 15–24, ⩾25 cigarettes per day), duration of smoking (<30, 30–39, ⩾40 years), total alcohol drinking (<2, 2 to <4, 4 to <8, ⩾8 drinks per day), duration of alcohol (<30, 30–39, ⩾40 years) and nonalcohol energy intake (quintiles) (Breslow and Day, 1980 ).
We also estimated the ORs for combinations of selected food groups (i.e., fruit, vegetable and meat) and lifestyle habits (i.e., tobacco smoking and alcohol drinking), and we tested the corresponding interaction by likelihood ratio tests.
Trained personnel interviewed both cases and controls during their hospital stay using a structured questionnaire, including information on sociodemographic characteristics, anthropometric measures and selected lifestyle habits (including tobacco smoking and alcohol drinking). Subjects' dietary habits during the 2 years before cancer diagnosis or hospitalisation (for controls) were assessed through a valid (Decarli et al, 1996 (link)) and reproducible (Franceschi et al, 1993 (link), 1995 (link)) food frequency questionnaire (FFQ), including information on weekly consumption of 78 foods, recipes and beverages. For a few vegetables and fruits, seasonal consumptions and corresponding durations were elicited. Food items were combined into 18 food groups: milk and yoghurt; cereals; soups; eggs; poultry; red meat; processed meat; fish; cheese; raw vegetables; cooked vegetables; all vegetables; potatoes; citrus fruits; other fruits; all fruits; desserts; and sugars. To estimate the daily intake of nutrients and total energy, we used an Italian food composition database (Gnagnarella et al, 2004 ). To evaluate the role of macronutrients independently from total energy intake, we derived energy-adjusted nutrients according to the residual method (Willett and Stampfer, 1986 (link)).
We categorised food groups, micronutrients and energy-adjusted macronutrients into quintiles (or quartiles/tertiles for a few foods with low frequency of consumption), according to the distribution among the control population. The ORs and corresponding confidence intervals for quantiles of intakes were estimated using multiple logistic regression models, including terms for age (5-year groups), sex, centre (Italy, Switzerland), education (<7, 7 to <12, ⩾12 years), year of interview (continuous), body mass index (BMI, <20, 20 to <25, 25 to <30 kg m−2), tobacco smoking (never smoker, ex-smoker, current smoker of: <15, 15–24, ⩾25 cigarettes per day), duration of smoking (<30, 30–39, ⩾40 years), total alcohol drinking (<2, 2 to <4, 4 to <8, ⩾8 drinks per day), duration of alcohol (<30, 30–39, ⩾40 years) and nonalcohol energy intake (quintiles) (Breslow and Day, 1980 ).
We also estimated the ORs for combinations of selected food groups (i.e., fruit, vegetable and meat) and lifestyle habits (i.e., tobacco smoking and alcohol drinking), and we tested the corresponding interaction by likelihood ratio tests.
