ABC | Volume 113, Nº4, October 2019

Original Article Francisco et al. Smoking and unhealthy diet Arq Bras Cardiol. 2019; 113(4):699-709 these factors is seen at more advanced ages. However, early signs of changes in health status occur more frequently from 40 years of age. 18,19 Brazil has an adult population (18 to 59 years of age) of approximately 114 million. The co-occurrence of smoking and an unhealthy diet has been under-investigated in the literature. Exposure to behavioral risk factors begins early in life 16,18 and is consolidated in adulthood, 13 with effects on health in different phases of life. Therefore, the objective of this study was to estimate the co-occurrence of smoking and unhealthy eating practices in the Brazilian adult population as well as determine associations with socio-demographic characteristics and health indicators. Methods A cross-sectional population-based study was conductedwith a sample of adults (18 to 59 years old) residing in the capitals of the 26 states of Brazil and the Federal District. The data were extracted from the records of 28,950 individuals interviewed in Sistema de Vigilância por Inquérito Telefônico (Vigitel [Brazilian Health Surveillance Telephone Survey]), in 2014. A minimum sample of 1,500 individuals in each city was established to estimate the frequency of any risk factor in the adult population 20 considering a 95% confidence interval and a maximum error of three percentage points. 7 Data collection was performed in three steps. The first step consisted of systematic random selection of at least five thousand telephone lines. This systematic selection stratified by postal code was performed using records of residential landlines registered with telephone companies. The lines selected in each city were submitted to a second random selection divided into replicates of 200 lines, with each replicate reproducing the same proportion of lines per postal code of the original registry. The third step was the random selection of one of the adults residing in the selected homes (after identification) among the lines considered eligible for the system. The following were excluded in this step: business lines, out-of-service or nonexistent lines and lines for which there was no answer after six calls on different days and at different times, including weekends and evening hours 7 (Figure 1). Weighting factors were used to compensate for the bias of the non-universal coverage of landlines. Using a post‑stratification weight calculated based on 36 analysis categories by sex (female and male), age group (18–24, 25–34, 35–44, 45–54, 55–64 and ≥65 years old) and level of education (none or incomplete primary school, complete primary school or incomplete high school, complete high school or incomplete university degree and complete university degree), the estimates were adjusted to the population. The “rake” method was used for the calculation Figure 1 – Flowchart of sample selection process. Vigitel, 2014. 1 st step 2 st step 3 st step • Selection of residential landlines – Systematic random selection stratified by postal code, based on electronic records from OI, GVT, Telefônica and Embratel/Claro telecommunication companies; – 5,000 lines randomly selected from each of 26 state capitals of Brazil and Federal District (n = 135,000); – Sample planned to reach at least of 1,500 interviews per city. • Organization of selected lines into replicates – In each city, selected lines were submitted to new random selection and divided into 25 replicates (200 lines each) due to difficulty in previously estimating proportion of active residential lines; – 101,200 telephone calls made, distributed among 506 replicates; – Exclusion of non-eligible lines; – Identification of 62,786 lines eligible for system. • Selection of participants from selected homes – Random selection of one adult (≥ 18 years) from selected homes; – 40,853 interviews obtained. Vigitel 2014 700