Simultaneity and aggregation of risk factors for noncommunicable diseases among brazilian adolescents

Objective: To estimate the prevalence of concurrency of risk factors for Noncommunicable Diseases and non-random aggregation of these in Brazilian school adolescents. Method: Descriptive study, with data from the National School Health Survey (Pesquisa Nacional de Saúde do Escolar), 2015. The prevalence of concurrent risk factors was estimated: smoking, alcohol use, physical inactivity and low consumption of fruits and vegetables. Aggregation was verified when the observed-expected prevalence ratio was greater than 1. The analyses were performed in Stata 15.0 software. Results: The proportion of concurrent risk factors was higher for those who had two risk factors (56.1%; 95%CI: 55.5-56.6). The most prevalent risk factors combination was “Low consumption of fruits and vegetables + Physical inactivity” (66%; 95% CI: 65.8-66.9). There was an aggregation of risk factors, and the highest ratio between observed and expected prevalence was: “Smoking + Alcohol” (79.0%; 95%CI: 73.8-84.2). Conclusion and implications for practice: The findings show a high prevalence of concurrent risk factors, with aggregation of these in adolescents. Identifying these risk groups may favor early interventions by minimizing exposure and targeting primary prevention strategies in early exposure to risk factors.


INTRODUCTION
Noncommunicable Diseases (NCD) account for 41 million deaths annually, which is equivalent to 71% of all deaths worldwide. Cardiovascular diseases accounted for the majority of deaths from NCDs in 2015, followed by cancer, respiratory diseases and diabetes. These four disease groups account for more than 80% of all premature deaths by NCD 1 and share four behavioral and modifiable risk factors: inadequate diet, physical inactivity, alcohol and tobacco use 2 .
It is noted that these risk factors usually occur simultaneously 3 and evidence shows that involvement in multiple risk behaviors is associated with increased risk of chronic disease and mortality compared to involvement in one or no risk behavior 4 .
Currently, health agendas prioritize interventions targeting common risk factors in adults, stimulated by estimates that show that 70% of premature deaths in adults are largely due to behaviors started in adolescence, such as smoking, obesity, low levels of physical activity and alcohol consumption 5 .
Most studies on the subject treat these risk factors in isolation [6][7][8][9][10][11] . Research on the simultaneity of risk factors for NCDs, with only adolescents, is scarce and recent in Brazil, having investigated beyond the prevalence of simultaneous risk factors 12 , aggregation 13,14 and associated factors 14,15 . Although they bring relevant results, these are local studies, not having national representativeness, justifying the relevance of this study.
Results from previous editions of the National School Health Survey (PeNSE -Pesquisa Nacional de Saúde do Escolar) show prevalence of risk behaviors in adolescents in isolation16. Nevertheless, there is still no data on the simultaneous prevalence of these risk factors with a representative sample of Brazilian school adolescents.
Estimating the prevalence and non-random aggregation of concurrent risk factors may be useful for health managers and policy makers to recognize groups with similar characteristics based on patterns of risk behavior, targeting prevention strategies, reducing early risk exposure. The purpose of this study was to estimate the prevalence of the risk factors for NCD and the non-random aggregation of these factors in Brazilian school adolescents using data from PeNSE 2015.

METHOD
This is a cross-sectional, descriptive epidemiological study with secondary data from the 3 rd edition of PeNSE 2015, available on the website of the Brazilian Institute of Geography and Statistics (IBGE -Instituto Brasileiro de Geografia e Estatística) (http://www.ibge.gov.br). The target population was school adolescents who attended the 9th year of elementary school and students aged 13 to 17 years who attended the 6th grade of elementary school through the third year of high school, enrolled in the academic year 2015 and regularly attending public schools and private areas located in urban and rural areas throughout the National Territory 17,18 .
The sampling process was done by conglomerates (complex plans), using three stages of selection: in the first one were selected the municipalities or groups of municipalities (Primary Sampling Unit -PSU); in the second, the schools (Secondary Sampling Unit -SSU); and in the third, the classes (Tertiary Sampling Unit -TSU). All the students present on the day of collection in the drawn classes were invited to participate in the research. The sizing of the sample size in each stratum considered a maximum sampling error of approximately 3%, to estimate a 50% ratio, with a confidence level of 95% 17 .
Planning resulted in two samples, with different sampling plans. The sample that contemplates the students attending the 9 th grade (old 8 th grade) of Elementary School was denominated Sample 1; and the one composed of students from 13 to 17 years of age, attending the 6 th to 9 th year of Elementary School and the 1 st to 3 rd grade of High School was denominated Sample 2. Sampling was performed so that it could generate population estimates for both samples, in the geographic strata considered. Details of the sampling process, specificities of the two samples and calculation of the PeNSE 2015 sample weights are available in the complete PeNSE 2015 report 17 . The data were collected after contact and dissemination between the State and Municipal Education Offices, with the direction of the selected schools in each municipality and with the parents of the students of the classes chosen to clarify the objectives and procedures of the study.
In this study, the data referring to Sample 1 of the PeNSE 2015, which had a final sample of 102,072 students, were analyzed. The variables used were: children who reported drinking a glass or at least one dose of alcoholic beverage at least once in the 30 days prior to the survey, categorized as: "No, none" or "Yes, one or more in the last 30 days") 19 ; 6. Tobacco use in the last 30 days (percentage of school adolescents who reported smoking at least once in the 30 days prior to the survey, regardless of frequency and intensity, categorized as: "No, never smoked, no day" or "Yes, one or more days in the last 30 days") 19 ; 7. Low consumption of fruits and vegetables (percentage of school adolescents who consumed less than five servings of fruits and vegetables in the last seven days prior to the survey) 20 ;

Simultaneity and aggregation of risk factors
Rocha FL, Velasquez-Melendez G 8. Physical inactivity, as measured by overall estimated physical activity (The overall estimated physical activity refers to the number of days that the students declared to do at least one hour per day of physical activity in the seven days prior to the survey, with the estimated time in minutes were considered as those who counted a time ≥ 300 minutes and inactive those who counted a time <300 minutes) 17,19 .
The dependent variable "Simultaneous risk factor" was generated by the sum of scores of the four factors (alcohol, tobacco use, physical inactivity and inappropriate intake of fruits and vegetables) ranging from 0 to 4, where "0" represents absence of any of these factors and the other values represent the number of factors simultaneously present. Each variable was categorized as "0" to indicate absence and "1" to indicate presence.
Statistical analyzes were performed in Stata 15.0 software, considering the complexity of the sampling plan. For this, we used the different weights of the observations that influence the point estimates of parameters of the total population by survey command. The individual proportions of risk factors were stratified by gender and estimates were produced for the population. Pearson's Chi-square test was used to evaluate the difference between groups. A 95% Confidence Interval (95% CI) and a significance level of 5%.
The observed prevalence of the 16 possible combinations of risk factors were estimated. The non-random aggregation of risk factors was verified when the Observed Prevalence (OP) of the combination of factors exceeded the Expected Prevalence (EP). EP of a specific pattern of risk factor aggregation was calculated by multiplying the individual probabilities of each risk factor occurring in the sample. Therefore, it was possible to investigate which combinations were higher or lower than the expected values, assuming that the risk factors occur independently in the study population. The result between the OP ratio for EP (OP/EP) greater than 1 indicates the existence of non-random factors 21

RESULTS
This study analyzed data from Sample 1 of PeNSE 2015 composed of 102,072 individuals. For the purposes of this analysis, a subpopulation was used, considering the observations whose variables of interest for this study had no missing data. Thus, this sample is composed of 99,738 individuals, the majority being female (51.6%; CI: 50.9-52.2), with 14  3) did not know inform mother's schooling ( Table 1).
The proportion of risk factors isolated was higher for low consumption of fruits and vegetables (80.6%; 95% CI: 80.1-81.2) and physical inactivity (79.5%; 95% CI: 79.0 -80.0), followed by alcohol consumption (23.8%, 95% CI: 23.2-24.4). The lowest risk factor was tobacco consumption (5.6%, 95% CI: 5.2-5.9). There was a significant difference between genders for alcohol use (22.5% for males and 25.1% for females, p <0.0001) and physical inactivity (71.7% for males and 86.9% for the female p <0.0001) ( Table 1). Figure 1 shows the proportion of concurrent risk factors: alcohol, tobacco, physical inactivity and low consumption of fruits and vegetables for the total sample and stratified by gender. The highest proportion was two simultaneous risk factors (56.1%, 95%CI: 55.5-56.6). When comparing by gender, the proportions of two (58.5; 95%CI: 57.7-59.3), three (16.6%, 95% CI: 16.0-17.2) and four (3.2 (95% CI: 2.9-3.6). Simultaneous risk factors were higher for females, with significant difference between groups (p <0.0001). Figure 2 shows the proportions of simultaneous risk factors by geographical regions of Brazil among female and male adolescents. It was observed that the highest proportions found among the female were for those who had up to two simultaneous risk factors in the five regions, being the proportion of two risk factors in the Northeast (62.7%) and North (62.5%), respectively. Among adolescents with three and four risk factors, the highest proportions were found in the South (22.7% for 3 and 4.4% for four risk factors) and in the Southeast (17.1% for three and 3, 6% for four simultaneous risk factors).
The highest proportions found among male adolescents were also for those who had two risk factors. However, when comparing by gender, in all regions, the proportions of two and three risk factors were lower among male adolescents, when compared to the female. Among adolescents with four risk factors, the proportions of the Northeast region were similar in both genders. In the North region it was slightly higher among men and, in the other regions, the proportions among the male adolescents were lower than the proportions found among female adolescents. Table 2 describes the observed and expected prevalence of different combinations of risk factors for NCDs and their aggregations in adolescents. The most prevalent combination was "Low consumption of fruits and vegetables + Physical inactivity" (OP: 66%, 95%CI: 65.8-66.9). All risk factors presented agglomeration, except for "Low consumption of fruits and vegetables + Physical inactivity + Alcohol" (OP/EP: 1.1; 95%CI: 1.0-1.1). It should be noted that the highest prevalence ratios Simultaneity and aggregation of risk factors Rocha FL, Velasquez-Melendez G

DISCUSSION
This study showed the isolated and simultaneous occurrence of the four main risk factors for NCD and the aggregation of these factors in a representative sample of Brazilian school adolescents, with data from the PeNSE 2015.

Simultaneity and aggregation of risk factors
Rocha FL, Velasquez-Melendez G  The proportion of risk factors isolated, such as low consumption of fruits and vegetables and physical inactivity, was high. When examining the proportion of multiple risk factors, more than half had at least two factors. The most prevalent risk factors combination was the low consumption of fruits and vegetables with physical inactivity. Only 4.7% of adolescents did not present any factor.
Comparability with other similar studies is restricted, since the inclusion of risk factors was varied, including the fact that these studies are of local scope. International surveys on the simultaneity of risk factors in different population groups were carried out using population surveys, whose indicators used were the same as in this study: smoking, alcohol consumption, physical inactivity and poor diet in fruits and vegetables. These studies found the following prevalence for two or more concurrent risk factors: 68% in people aged 16 years or older in England; 34% in men and 26% in women aged 25 years or more in Switzerland 23 and 39.2% in men; and 24.6% in adult women in Finland 24 .
Regional differences were also observed in this study, and in the North and Northeast regions, greater proportions were found for those who had two risk factors in both genders. The adolescents with the highest proportions of three and four risk factors resided in the South and Southeast regions. A metaanalysis study showed that the prevalence of two and three simultaneous risk factors in adolescents in middle-and lowincome countries worldwide was 40% and 10%, respectively, and varied according to geographic region studied 19 . Authors failed to explain the regional differences found in the study but believe that social norms and cultures can shape adolescents' health behavior and political and economic factors can influence those choices, and it is relevant to explain these differences 19 .
The high prevalence of adolescents with two or more simultaneous risk factors described in our results has important implications, considering that evidence from longitudinal studies on health risk behaviors shows changes in behaviors related to physical activity, fruit intake, smoking and drunkenness according to the prevailing social norms about these behaviors at different ages, from adolescence to adulthood 25 . It should be emphasized that the behaviors can vary throughout the adolescence, being able to be a mix of healthy and unhealthy behaviors 26 .
The highest prevalence of two, three and four risk factors found among women were reported in a study with adults. However, it cannot be said that women would be at greater risk than men because there is no clear explanation for these differences, which requires further investigation. It should be noted that many of the studies on concurrent risk factors use different risk indicators, analytical techniques and cut-off points and are therefore inconclusive about which group would be at greatest risk 21 .
The most prevalent combination in this study was "Low consumption of fruits and vegetables + Physical inactivity". A systematic review study synthesized evidence from different countries and showed that the simultaneous occurrence of risk factors and the most common combinations found in adolescents differ from those found in adults 27 .
Findings also showed that the most prevalent aggregations were "Smoking + Alcohol", "Low consumption of fruits and vegetables + Smoking + Alcohol" and "Physical inactivity + Smoking + Alcohol". There is evidence that in adults the use of alcohol and tobacco tends to be added 27 . In England, we found aggregations of "Smoking, alcohol consumption and inadequate consumption of fruits and vegetables" and "Physical inactivity and inadequate consumption of fruits and vegetables" in a population older than 16 years 21 . In adolescents, studies showed associations between inadequate fruit and vegetable intake and low physical activity, but did not show aggregation patterns, since they were not the focus of investigations 28,29 . Thus, there is a limitation in the comparability with studies conducted in other countries to verify the existence of an association of these same risk factors in adolescents.

Simultaneity and aggregation of risk factors
Rocha FL, Velasquez-Melendez G It is noteworthy that in all the most prevalent risk factor aggregations, smoking and alcohol were present. Results from the Pesquisa Nacional Longitudinal de Crianças e Jovens (greely translated as National Longitudinal Children and Adolescents Survey), with a representative sample of Canadian children and adolescents, showed aggregation of tobacco and alcohol and, except for the combination of physical inactivity, sedentary behavior and high Body Mass Index (BMI), all combinations of three and four behaviors that were included smoking and alcohol 30 . Although authors have worked with different risk factors (physical inactivity, sedentary behavior, smoking, alcohol and high BMI), the results corroborate our findings about the importance of the addition of "Smoking + Alcohol".
As the use of alcohol is socially accepted, many adolescents start using it early and remain in use in adulthood. There is a positive relationship between alcohol and tobacco, suggesting that alcohol stimulates tobacco use. It is believed that this is because these behaviors are more influenced by cultural aspects 31 . This may not be enough, but explain in part the fact that the observed prevalence of the combination "Smoking + Alcohol" was much higher than expected when compared to other combinations.
Among the limitations of this study, we cite the bias of social desirability, since adolescents can provide answers that underestimate the prevalence of risk behaviors, aiming to show that they adopt socially accepted behaviors. These results have as a strong point: rigor of the methods used in PeNSE, the fact of having a minimum non-response rate and its national representativeness.

CONCLUSIONS
The results found in this study show a high prevalence of simultaneous risk factors in Brazilian adolescents and, therefore, indicate the need for preventive actions with strategies that include behavioral changes throughout the life cycle, since behaviors are initiated in the childhood and may persist in adolescence and adulthood.
This study has important implications for the practice, since knowing the prevalence of the main risk factors for NCDs and their clusters in adolescents may direct primary prevention strategies, behavior change and the promotion of a healthy transition between the adolescent and adulthood phase.

Simultaneity and aggregation of risk factors
Rocha FL, Velasquez-Melendez G