Trends in sociodemographic and lifestyle factors associated with sedentary behavior among Brazilian adults

Objective: To analyze sociodemographic and lifestyle factors associated with screenbased sedentary behavior (watching television ≥3 hours/day) among adult individuals in Brazil. Methods: Quantitative analysis of ten editions of the cross-sectional health survey VIGITEL, representative at population level. Individuals from states’ capitals living in households with land-line telephone were randomly selected and interviewed using structured questionnaire by telephone. A multivariate logistic regression model was estimated for identification of factors associated with screen-based sedentary behavior. Results: There was stability in trends referring to prevalence of sedentary behavior from 2008 to 2017. Prevalence of sedentary behavior was higher between individuals with unhealthier lifestyles: consumption of <2 in natura food items (vegetables, fruits and beans) per day (26.73% [95%CI 25.2%;28.31%]) in comparison to ≥2 items per day (23.79% [95%CI 21.92%;25.77%]); consumption of soft drinks ≥5 days per week (31.24% [95%CI 29.58%;32.95%]) than <5 days per week (23.82% [95%CI 22.2%;25.52%]); and practice of <150 minutes of physical activity per week (28.2% [95%CI 26.17%;30.33%]) than ≥150 minutes per week (22.54% [21.27%;23.86%]). Regular consumption of in natura food items (OR=0.984), practice of physical activity (OR=0.798) and living in richer municipality (OR=0.826) represented protective factors in relation to screen-based sedentary behavior, whilst regular consumption of soft drinks (OR=1.440), smoking (OR=1.375) and alcohol abuse (OR=1.334) represented risk factors. Conclusion: The adoption of screen-based sedentary behavior among adult individuals in Brazil presented significant association with modifiable behavioral factors in the period 2008-2017.

Evidence on associations between sedentary behavior and chronic non-communicable diseases indicates significant association with cardiometabolic 3 and cardiovascular 4 diseases, cancer 5 , overweight and obesity 6 , and overall mortality 7 .
The recently published guidelines of the World Health Organization on physical activity and sedentary behavior recommend that adult individuals (18-64 years-old) limit sedentarism, especially by replacing sedentary activities with physical activity at least 150 to 300 minutes per week due to substantial benefits to individuals' health, which contributes to well-being and overall quality of life 8,9 .
Despite the harmful effects of sedentarism on health status, it presents high prevalence in diverse countries worldwide 8 . A recent study indicated that approximately 65% of adults in the United States devoted two or more hours per day watching television in 2015-2016 10 .
It is important to emphasize that sedentary behavior and physical activity may be performed in different domains, e.g., during leisure, transportation or labor, and other occupational or educational activities. In general, sedentary behaviors that occur during leisure are considered discretionary, and time spent watching television is usually adopted as proxy variable for optional sedentary behavior in epidemiological studies, especially considering its sensitivity to influences from cultural and socioeconomic contexts [11][12][13] .
Although there is emerging academic interest in factors associated with sedentary behaviors, most studies focus on high-income countries 12,13 , and there is lack of population level evidence for low-and middle-income countries, e.g. Brazil, especially

Study design
The study presents analysis of datasets from the Surveillance of Risk and Protection Factors for Prevention of Chronic Diseases through Telephone Survey (VIGITEL), conducted by the Brazilian Ministry of Health, including ten years of cross-sectional observational individual-level data from representative sample of the adult population living in Brazilian state capitals and Federal District in the period from 2008 to 2017.

Databases
The VIGITEL is a telephone health survey conducted annually since 2006 by the Brazilian Ministry of Health to monitor risk and protection factors for chronic diseases in the Brazilian population. The databases include individual-level information for each year of the survey, available at the Brazilian Ministry of Health website. Microdata from surveys conducted from 2008 onwards were selected, considering the consolidation process of the survey during the first two years after implementation.
VIGITEL sampling process is based on minimum sample of 1,500 individuals from each of the Brazilian state capitals and Federal District for estimation of frequency of risk and protection factors for chronic diseases in the adult population with 95% confidence and maximum error of three percentage points 14 .
The first stage of sampling refers to random selection of at least 5,000 landlines per municipality from landline registrations of main telephone companies in the country.
After initial drawing, lines eligible for survey are selected, i.e., active residential lines.
The second stage of sampling consists of drawing one adult per household to participate in the survey 14 . Considering the survey sample design, individuals interviewed are assigned weights to allow statistical inferences in relation to the population of 26 state SciELO Preprints -This document is a preprint and its current status is available at: https://doi. org/10.1590/1980-549720210014.supl.1 capitals and Brazilian Federal District, using rake method 14 . Data collection was carried out through structured interview with application of closed questionnaire by telephone 14 .
In addition to information from VIGITEL, data referring to Gross Domestic Product (GDP) and population of each municipality, obtained from the Brazilian Institute for Geography and Statistics (IBGE), were included in the dataset to represent certain environmental aspects of the municipality and population economic status and to assess potential effects of economic conjuncture on other variables in the survey period (2008 to 2017).

Variables
Sedentary behavior (outcome) was based on self-reported daily time watching television, considering sedentary individuals with screen time equal or higher than 3 hours per day.
Variables of interest in the present study were: (1) self-reported frequency of consumption of in natura food items (vegetables, fruits and beans) per week 15 ; (2) selfreported frequency of consumption of soft drinks per week 15 ; (3) sociodemographic characteristics: age, biological sex, educational attainment, ethnicity/skin color, marital status and occupation; (4) health characteristics: self-assessment of health status, selfreported diagnosis of diabetes, self-reported diagnosis of hypertension, overweight and obesity; (5) self-reported behavioral characteristics: physical activity, alcohol abuse, and smoking; (6) GDP per capita in the municipality of residence of the individuals, using data obtained from IBGE.

Data processing
SciELO Preprints -This document is a preprint and its current status is available at: https://doi. org/10.1590/1980-549720210014.supl.1 Information of VIGITEL databases ranging from 2008 to 2017 were further organized into a single dataset, after selection of variables compatible along the period of analysis, to allow statistical analysis on trends and factors associated with adoption of screenbased sedentary behavior among adult individuals.
A set of variables from VIGITEL was converted into binary variables, coded into 0 (no) and 1 (yes) values, according to specific criteria based on evidence of the literature or cutoff points established by national and/or international organizations: screen-based sedentary behavior, and regular consumption of in natura food items and soft drinks.
The adoption of screen-based sedentary behavior was based on self-report of daily duration of television watching, considering the cutoff point of three or more hours per day.
Regular consumption of in natura food items and soft drinks was based on self-reported frequency per item: none day, 1 to 2 days, 3 to 4 days, 5 to 6 days or all days of the week 14 . Three variables registering self-reported frequency of in natura food items consumption, i.e., beans, fruits and vegetables (considered markers of healthier food consumption patterns), were converted into number of days per week consuming each item, which were added up and divided by seven days per week to comprise total in natura food items consumed per day. Then, it was categorized using cutoff point of at least two items per day during the week 14 .
Regular consumption of soft drinks (considered marker of unhealthier food consumption patterns) was categorized using cutoff point of consuming five or more days during the week 15 .
Regarding sociodemographic characteristics, age and educational attainment were continuous variables maintained in its original format for the analysis. Biological sex, ethnicity/skin color, marital status and occupation were converted into categorical SciELO Preprints -This document is a preprint and its current status is available at: https://doi. org/10.1590/1980-549720210014.supl.1 variables, encompassing the following categories, respectively: female (0) and male (1); white (0) and black, brown and indigenous (1); living with companionship, i.e., married and stable union (0) and living without companionship, i.e., single, divorced and widowed (1); and currently working (1) and not working (0).
Amongst health characteristics, self-assessment of health status in five categories (very good, good, fair, poor or very poor) was converted into binary variable considering individuals who declared having poor or very poor health status. Presence of diabetes or hypertension were registered according to self-report of the individual. Occurrence of overweight (BMI ≥ 25 kg/m²) 14 and obesity (BMI ≥ 30 kg/m²) 14 in the VIGITEL was based on the estimation of the Body Mass Index (BMI), based on self-reported information about weight and height 14 .
Behavioral characteristics were adopted in its original format from VIGITEL: physical activity level (≥ 150 minutes per week), alcohol abuse (≥ 5 doses for men; ≥ 4 doses for women at least on one occasion in the last 30 days), and smoking (current use of tobacco products, regardless the amount) 14 .
Values of GDP per capita were updated by applying the National Consumer Price Index year of survey, and cross-effects of municipality and year of survey, in order to capture SciELO Preprints -This document is a preprint and its current status is available at: https://doi.org/10.1590/1980-549720210014.supl.1 potential influence of local policies. Analyses were performed using statistical software Stata® (Stata Corp., College Station, USA), version 14.2 for Windows, applying svyset command for sample design using rake weighting method, considering statistical significance p0.05.

Ethical aspects
The VIGITEL survey project was approved by the National Commission on Research Ethics (CAAE: 65610017.1.0000.0008). Informed consent was obtained verbally at the time of telephone contact 14 .

RESULTS
Participants in VIGITEL survey were usually female individuals, individuals who declared themselves black, brown or indigenous, and individuals who worked. The proportion of young adults (18 to 39 years-old) was higher during the first survey editions; nevertheless, there was an increasing trend in participation of older adults (40 to 59 years-old), and elderly individuals (over 60 years-old) (  1 and table S1).
SciELO Preprints -This document is a preprint and its current status is available at: https://doi.org/10. 1590/1980-549720210014.supl.1 There were no statistically significant differences among individuals interviewed throughout the period regarding biological sex, occupation, self-assessment of poor health status and self-reported hypertension diagnosis (table 1 and table S1). In Brazil, evidence referring to the protective role of regular in natura food consumption and recommended levels of physical activity in relation to adoption of sedentary behavior was also observed in previous study performed among public school teachers in Presidente Prudente municipality, state of Sao Paulo 20 ; in addition, there were also associations between overfeeding and alcohol consumption with time spent watching television 20 .

Results
Results obtained in the study, reinforced by evidence in previous studies, draw attention to the concomitance and the repercussion of harmful habits to individual's health, indicating the importance of the discussion on the presence of multiple behavioral risk factors in relation its impacts on health outcomes. Adherence to healthy lifestyles, i.e., combination of healthier behaviors, was significantly associated with reduction in SciELO Preprints -This document is a preprint and its current status is available at: https://doi. org/10.1590/1980-549720210014.supl.1 premature death in the United States, resulting in increase in life expectancy, particularly healthy life years free from chronic non-communicable diseases (NCD) 21,22 . Similar evidence has been observed in studies with Brazilian adolescents 23 and Polish adults 24 , particularly regarding eating patterns associated with screen-based sedentary behavior (including watching television). It points out to underlying mechanisms of encouragement for consumption of food items that are considered markers of unhealthier food consumption patterns (e.g., soda, snacks, and sweets) whilst watching television. Furthermore, the habit of watching television has been related to body fat deposits 24 , increasing risks of overweight, abdominal obesity, higher BMI and waist circumference 6,25,26 , which were partially observed in the results of the study also.
It is important to emphasize that overweight and obesity are important risk factors for NCD 27,28 , being responsible for substantial health and economic burden in populations, health systems and households worldwide, considering direct costs with treatments and indirect costs for individuals, e.g., productivity losses, family commitment and emotional health impacts 29-31 .
Our results have shown that, besides overweight and obesity, individuals who selfreported diabetes and hypertension diagnosis were also more likely to adopt screenbased sedentary behavior, association also observed in previous studies 3,32 . The evidence on the relationship between NCD and sedentary behavior fosters the discussion on the need for engagement of individuals diagnosed with NCD within initiatives for promotion of physical activity. It is especially important regarding perceived barriers to achieve recommended physical activity practice 33,34 , and adoption of healthier eating patterns 35,36 , which highlight the social and environmental influences on behavioral change.
SciELO Preprints -This document is a preprint and its current status is available at: https://doi. org/10.1590/1980-549720210014.supl.1 In the context of sociodemographic characteristics, results of the study showed higher tendency of screen-based sedentary behavior among individuals who declared being single, divorced, or widower; i.e., individuals living without companionship, in accordance with previous studies with Canadian and Japanese adults 37,38 .
However, a systematic review has shown certain inconsistencies regarding the influence of family and household factors, including marital status, on adoption of sedentary behavior during leisure 12 . Therefore, although some evidence point to the adoption of screen-based sedentary behavior among individuals living unaccompanied, additional research is required to identify whether marital status influences sedentary habits like watching television, and its relationship with other sociodemographic factors over time.
In any case, evidence call for attention towards the discussions about influences of peers in adoption of healthier lifestyles, encouraging and/or accompanying practice of physical activity during leisure, instead of sedentary recreational activities like watching television 37,38 .
In terms of ethnicity/skin color, it is important to point out that higher trends in screenbased sedentary behaviors among individuals declaring themselves black, brown or indigenous may be linked to environmental characteristics that impose barriers to physical activity practice in ethnic minorities, according to evidence from studies in United Kingdom 39 and United States 40 . Thus, it represents an opportunity to discuss design of health policy interventions with an equitable orientation, focusing on specific characteristics of the Brazilian black, brown, or indigenous individuals.
Sedentary behavior presents socioeconomic and cultural determinants related to the organization of contemporary society, labor and educational activities, i.e., routines that have been designed to occur generally in sitting position with minor energy expenditure, promoting sedentarism in individuals and populations 11 . Whilst adherence to physical SciELO Preprints -This document is a preprint and its current status is available at: https://doi.org/10. 1590/1980-549720210014.supl.1 activity is commonly associated with leisure in high-income countries and work in lowincome countries, it is possible to observe both situations in middle-income countries like Brazil 41 . Therefore, the adoption of indicators like watching television three or more hours per day for analysis of sedentary behavior may comprise an important marker of discretionary recreational activity, unlike other forms of screen-based sedentary behavior, e.g., duration of activities using computer, which may be linked to occupational activities. Our results showed that individuals who declared that were not working presented higher tendency to maintain screen-based sedentary behavior during leisure by watching television ≥3 hours per day. In addition, changes in the survey questionnaires throughout the period of analysis limited the possibility of including certain characteristics of interest in the study, like presence of hypercholesterolemia, consumption of other food items (milk, meat, and sweets), among others. Therefore, only variables that remained directly comparable during the period analyzed were selected in the study, allowing consistency for estimation of the model proposed.
SciELO Preprints -This document is a preprint and its current status is available at: https://doi.org/10.1590/1980-549720210014.supl.1 VIGITEL includes self-reported characteristics through telephone survey, which may result in underestimation of characteristics that individuals' believe "wrong" or "socially unacceptable" and overestimation of characteristics perceived "right" or "socially acceptable", thus reducing accuracy of analysis referring to certain individual's characteristics and behaviors. Furthermore, the variable for screen-based SciELO Preprints -This document is a preprint and its current status is available at: https://doi. org/10.1590/1980-549720210014.supl.1 Powered by TCPDF (www.tcpdf.org) Finally, increasing trends towards sedentarism, alcohol abuse, overweight and obesity during period analyzed represent a call for action within the context of the Brazilian health system, especially directed to primary health care strategies for health promotion and disease prevention. Considering the lack of cohort data representative at national level in Brazil, results of the study may subsidize the formulation of strategic interventions in public health policies to promote healthy lifestyles among Brazilian adults.
SciELO Preprints -This document is a preprint and its current status is available at: https://doi. org/10.1590/1980-549720210014.supl  Data presented in number of individuals, n (%). P-values obtained from Pearson's chi-square test among years of the survey.