ABC | Volume 113, Nº4, October 2019

Original Article Francisco et al. Smoking and unhealthy diet Arq Bras Cardiol. 2019; 113(4):699-709 Table 1 – Scoring scale for unhealthy consumption of foods. Vigitel, 2014 Foods 0 1 2 3 4 Beans Every day 5 to 6 days a week 3 to 4 days a week 1 to 2 days a week Never or hardly ever Fruits Every day 5 to 6 days a week 3 to 4 days a week 1 to 2 days a week Never or hardly ever Raw vegetables 1 Every day 5 to 6 days a week 3 to 4 days a week 1 to 2 days a week Never or hardly ever Cooked vegetables 2 Every day 5 to 6 days a week 3 to 4 days a week 1 to 2 days a week Never or hardly ever Milk Every day 5 to 6 days a week 3 to 4 days a week 1 to 2 days a week Never or hardly ever Red meat 3 Never or hardly ever 1 to 2 days a week 3 to 4 days a week 5 to 6 days a week Every day Sweetened soft drink or artificial juice Never orhardly ever 1 to 2 days a week 3 to 4 days a week 5 to 6 days a week Every day Sweets 4 Never or hardly ever 1 to 2 days a week 3 to 4 days a week 5 to 6 days a week Every day 1 Lettuce and tomato salad or salad with any other raw vegetable. 2 Consumption of vegetables cooked with food or in soup, such as collards, carrot, eggplant, zucchini, except potato, cassava or yam. 3 Red meat: beef, pork, goat. 4 Consumption of sweets, such as ice cream, chocolate, cakes, cookies, etc. of the post-stratification weight of each individual in the sample. Information on the sample design of the Vigitel survey, the data collection instruments and procedures used in the interviews is published elsewhere. 7 In the present study, the co-occurrence of smoking and an inadequate diet was considered the variable of interest. A smoker was considered any individual who answered affirmatively to the following question: “Do you currently smoke?,” irrespective of the number of cigarettes, frequency and duration. The indicator of an unhealthy diet was created from a set of foods that serve as markers of the intake profile associated with protection from chronic diseases (beans, fruits, milk, raw vegetables and cooked vegetables) and risk for chronic diseases (red meat, sweets and sweetened beverages). Scores ranging from zero to four were attributed depending on the food and intake frequency. Markers of the protection category ingested daily and those of the risk category never or rarely ingested were not scored (zero). Maximum of four points was attributed to protective foods never or rarely consumed and risk foods ingested daily (Table 1). The total score was determined by the sum of food items and ranged from 0 to 32 points, with higher scores indicating poorer dietary quality. This variable was then categorized considering distribution terciles. Individuals in the 2 nd and 3 rd terciles (14 or more points) were grouped together, creating a dichotomous variable for unhealthy eating (yes or no). Co-occurrence was determined by the simultaneous occurrence of both of these conditions (smoking and unhealthy diet). The following socio-demographic variables were considered: macro-region of the country (North, Northeast, Central West, South and Southeast), sex (male and female), age group (18 to 39 and ≥ 40 years old), skin color/ethnicity (white, black, yellow, brown and indigenous), marital status (with and without a spouse), education (0 to 8, 9 to 11 and 12 or more years of study) and having a private health insurance plan (yes or no). The following variables related to behavior and health status were considered: body mass index (BMI) (< 25 kg/m 2 , ≥ 25 to <30 kg/m 2 and≥30 kg/m 2 ), binge drinking [five or more drinks for men and four or more drinks for women on a single occasion in the previous 30 days (yes or no)], practice of physical activity (active or inactive) and self-rated health (very good/good, fair or poor/very poor). Weight and height were self-reported by the respondents. BMI was calculated for all records based on the imputation of the measures of weight and height using the “hot deck” method. 7 The following diseases were also considered: arterial hypertension, diabetes mellitus and dyslipidemia (all categorized as “yes or no”). Statistical analysis Descriptive analysis was performed for the characterization of the study population. Age (continuous variable) was expressed as meanand respective95%confidence interval. Categorical variables were expressed as relative frequency (percentage). Prevalence values were estimated for smoking, unhealthy eating and the co-occurrence of both of these variables of interest according to socio-demographic characteristics, other behavioral factors and health conditions. Associations were determined between the co-occurrence of the risk factors and variables selected using Pearson’s chi-square test with second-order correction (Rao & Scott), considering a 5% significance level. Next, prevalence ratios were estimated and adjusted for sex and age according to socio‑demographic characteristics, behavioral factors and health conditions. A hierarchical Poisson regression model was used considering two sets of variables: 1) socio-demographic and 2) behavioral/ health conditions. The variables from the first block were incorporated into the model. Those that remained significant after adjustments by the other variables on the same hierarchical level remained in the model, to which the second block of variables was incorporated. All variables with p-value < 0.05 after adjustments for variables on the same and higher hierarchical level remained in the final model. The analyses were performed using the Stata statistical package, version 12.0. The objectives of the survey were made clear to all individuals contacted by telephone and written consent was substituted with verbal consent. The Vigitel study received approval from the National Human Research Ethics Committee of the Brazilian Health Ministry (certificate number: 355.590, June 26, 2013). 701