Meta-analysis of the prevalence of physical inactivity among Brazilian adolescents

A systematic review and meta-analysis were carried out to investigate the prevalence of physical inactivity among adolescents in Brazil. We identified articles that had been published up to August 2010 and the search was conducted using six electronic databases. We did not enforce any search limitations. Forest plot-type graphs were generated using the prevalence of physical inactivity stratified by region and sex. Meta-regression models were fitted to identify possible sources of heterogeneity in the prevalence estimates. Of the 1,496 articles initially identified, 37 were deemed appropriate for the systematic review. Prevalence rates ranged from 2% to 80% for male and from 14% to 91% for female subgroups. The lowest prevalence rates of physical inactivity were found to be for the Southern region, whereas the highest rates were observed in the North-Northeast. The methods employed to evaluate physical inactivity in Brazilian adolescents also differed among the studies. This variation demonstrates the need for standardised and validated methods of measuring physical activity in epidemiological investigation. Motor Activity; Adolescent; Meta-Analysis Introduction Childhood and adolescence are extremely important periods for the development of a healthy lifestyle because the behaviors acquired throughout these stages tend to be perpetuated along the course of a lifetime 1. Health-risk behaviors, such as physical inactivity, low fruit consumption, alcoholic-beverage consumption and tobacco use are increasingly present in our society and are associated with the development of chronic non-communicable diseases (NCDs) 2,3. Moreover, NCDs are increasingly prevalent among children and adolescents and are the leading cause of morbidity and mortality in Brazil as well as worldwide. NCDs can have serious personal, social and financial impacts 4,5. Physical activity and healthy eating are two highpriority behaviors for the promotion of health and the prevention of NCDs 4. The most comprehensive study on physical activity was conducted between 2002 and 2003, with over 200,000 adults (18-69 years of age) from 51 countries (mostly of low to middle income), including Brazil 6. Although the benefits of physical activity in relation to health are well established in the literature 7, this study observed an average prevalence of physical inactivity of 18% (unweighed, i.e., for all countries), whereas the prevalence rates in Brazil were 25% and 30% for men and women, respectively 6. REVISÃO REVIEW Barufaldi LA et al. 1020 Cad. Saúde Pública, Rio de Janeiro, 28(6):1019-1032, jun, 2012 In Brazil, several studies have examined physical activity. However, most of these studies have involved specific populations, i.e., representative samples of municipalities or states, not nationwide surveys. According to a systematic review published in 2009, which described articles on the prevalence of physical activity (or inactivity) in Brazil, the number of studies on this topic has largely increased since 2000. Also, according to this review, most studies were conducted with adults. Among the studies that evaluated adolescents, the prevalences of physical inactivity were discrepant, ranging between 30% and 70% 8. Considering the importance of physical activity among adolescents and the dispersed information from Brazilian surveys, this article aims to describe the prevalence of physical inactivity among Brazilian adolescents through a systematic review and a meta-analysis of studies on this topic.


Introduction
Childhood and adolescence are extremely important periods for the development of a healthy lifestyle because the behaviors acquired throughout these stages tend to be perpetuated along the course of a lifetime 1 .Health-risk behaviors, such as physical inactivity, low fruit consumption, alcoholic-beverage consumption and tobacco use are increasingly present in our society and are associated with the development of chronic non-communicable diseases (NCDs) 2,3 .Moreover, NCDs are increasingly prevalent among children and adolescents and are the leading cause of morbidity and mortality in Brazil as well as worldwide.NCDs can have serious personal, social and financial impacts 4,5 .Physical activity and healthy eating are two highpriority behaviors for the promotion of health and the prevention of NCDs 4 .
The most comprehensive study on physical activity was conducted between 2002 and 2003, with over 200,000 adults (18-69 years of age) from 51 countries (mostly of low to middle income), including Brazil 6 .Although the benefits of physical activity in relation to health are well established in the literature 7 , this study observed an average prevalence of physical inactivity of 18% (unweighed, i.e., for all countries), whereas the prevalence rates in Brazil were 25% and 30% for men and women, respectively 6 .
In Brazil, several studies have examined physical activity.However, most of these studies have involved specific populations, i.e., representative samples of municipalities or states, not nationwide surveys.
According to a systematic review published in 2009, which described articles on the prevalence of physical activity (or inactivity) in Brazil, the number of studies on this topic has largely increased since 2000.Also, according to this review, most studies were conducted with adults.Among the studies that evaluated adolescents, the prevalences of physical inactivity were discrepant, ranging between 30% and 70% 8 .
Considering the importance of physical activity among adolescents and the dispersed information from Brazilian surveys, this article aims to describe the prevalence of physical inactivity among Brazilian adolescents through a systematic review and a meta-analysis of studies on this topic.
The descriptors used in the review process were selected after consulting the DeCs (Health Sciences Descriptors, BIREME) and MeSH (Medical Subject Headings, PubMed) terms.The search was conducted in Portuguese and/or English (depending upon the database) using blocks of three concepts: the first with terms related to physical activity ("physical activity", "exercise", "fitness", "motor activity", "physical inactivity" and "sedentarism"), the second with terms related to adolescence ("adolescent" and "young") and the third with terms related to Brazil ("Brazil", "Brazilian" and each state separately).We used the logical operator "OR" to combine the descriptors within each block and the logical operator "AND" to combine the blocks.We also used a resource for term truncations, when necessary.We did not use search limits for the data, language, studydesign or sample size.
The inclusion criteria were as follows: (a) the sample population included adolescents (10 to 20 years of age) and the data for adolescents were presented separately if other age groups were included, (b) an assessment of physical inactivity was conducted, (c) data collection was performed in Brazil.We excluded reviews or studies that exclusively included adolescents with specific health conditions (obesity, hypertension, diabetes, etc.).We chose not to include theses, dissertations and monographs because a systematic search for these works was not logistically feasible.
Articles were first selected by their title and abstract and then by reading the pre-selected articles in full.Each paper was reviewed and selected by two reviewers (L.A.B. and G.A.A.) and, when there was a disagreement, a third person was consulted (K.V.B.).In cases in which there was more than one publication using the data from the same study, the most comprehensive article was selected.
In addition to the prevalence of physical inactivity, we recorded information on the location and time period of the data collection, type of population, assessment method and classification of physical activity.
The heterogeneity between the studies was assessed using the I² statistic 9 .In order to obtain the summary measures, we used random-effect models due to the substantial heterogeneity of the results.In order to combine the prevalence measures, logit transformations were made initially to address the asymmetrical distribution.These prevalences were weighted by the inverse of the variance of the logit.Forest plot-type graphs were constructed using the prevalence of physical inactivity by region and by sex.Metaregression models were adjusted to identify possible sources of heterogeneity among the prevalences.The variables considered for this analysis were region, age, percentage by gender, type of population, sampling, sample size, method of evaluation and cut-off point.

Results
Figure 1 demonstrates the flowchart for the article selection process.Of the 1,496 articles initially identified (after removing duplicates), 37 were deemed eligible for the systematic review.The main objective of 18 of these articles was to analyse the prevalence of physical inactivity among adolescents.The remaining 19 articles had different primary goals although physical activity was a covariate in the study.Of the 37 studies, 35 were cross-sectional studies and two included phases of cohort studies, which had a cross-sectional analysis.
The characteristics of the studies are shown in Table 1.Most studies were conducted in the Southern region (19 studies), followed by the Southeastern and Northeastern regions (9 studies each).Only one study was performed in the Northern region.We did not find any studies that were carried out in the Central-Western region.
Most of the studies were based on schools with only six based on households and two on specific population groups (trainees from a university and adolescents from families observed at a Primary Care Unit as well as from hypertensive families).
The sample size ranged from 49 to 5,028 teenagers; 12 of the 37 studies were performed with samples sizes of up to 500 individuals, 11 with sample sizes between 500 and 1,000 adolescents and 14 with sample sizes of more than 1,000 individuals.
In all studies, physical inactivity was measured indirectly (i.e. based on information provided by individuals).Questionnaires were the most commonly used instruments.However, some studies used daily logs to obtain information.Although most studies used previously validated questionnaires, 14 studies used questionnaires unique to their investigation.
There was a large amount of diversity in the areas considered in the assessment of physical activity.Some studies considered only vigorous and/or moderate-intensity physical activities (during or outside of school), whereas other studies considered travel, leisure and sports activities as physical activities, excluding activities during school.Finally, some studies considered all travel, leisure and sports activities, whether during or outside of school, as physical activity.Some articles did not specify the areas considered in the analysis.
With respect to the cut-off points for classifying physical inactivity, the most commonly used cut-off was < 300 minutes per week of moderate to vigorous activity.However, the cut-off values varied widely between studies.
The overall prevalence of physical inactivity, as well as the prevalence stratified by sex, are shown in Table 2.There was a wide variation in these estimates: the lowest was 5.42% 10 , and the highest was 91% 11 .
The heterogeneity of the results expressed by the I² statistic was 99%.Prevalence rates ranged from 2% to 80% for male and from 14% to 91% for female subgroups.Among females, 69.7% of the prevalences were above 50% whereas, among men, 29.4% were above 50%.
When we evaluated the characteristics of the studies as a source of heterogeneity in the results, using meta-regression models, we observed that the region in which the study was conducted partially explained the inconsistency among prevalence rates for inactivity, both for males (p = 0.08) and females (p = 0.02).Figure 2 shows that the South has the lowest prevalence of inactivity, whereas the North and Northeast have the highest estimated prevalences.
The figures representing gender-specific data in the meta-analysis were constructed from 34 studies.One study was exclusively conducted in male adolescents, and, thus, female-specific data were gleaned from 33 studies.Of the 37 studies in the systematic review, three did not have any information on the prevalence of physical inactivity according to sex and are therefore not represented in Figure 2.
In the female group, we identified other possible sources of heterogeneity through there was a decline in the average prevalence found in studies using non-random sampling compared with studies with random sampling (by cluster, p = 0.02) and among studies with sample sizes over 1,000 adolescents (p = 0.07).
None of the other variables, for both of the overall population and stratified by gender, were associated with inter-study variations in the prevalence of physical inactivity.

Discussion
This systematic review showed a wide range in prevalence of physical inactivity, as well as in the methods used to assess it, in studies conducted among Brazilian adolescents.We observed that the majority of studies were conducted in the Southern region.Although we investigated the profile of physical inactivity by region, this was not sufficient to explain the heterogeneity of the findings.
Most studies found a higher prevalence of physical inactivity among women, similar to findings in international studies 12,13,14 .Researchers have reported that this gender difference is related to cultural, behavioral, psychological and maturation factors 13,15 .However, there have been no studies conducted in Brazil to assess the possible explanations for this difference.
Despite the progress in developing tools to assess physical activity, there are still many limitations in these methods, which are amplified among children and adolescents.The patterns of physical activity among young people can be influenced by cognitive, physiological and biomechanical changes that occur during development 16,17 .Physical activity among young people tends to be intermittent and of a more variable intensity than among adults, usually consisting of activities that involve less planning and less time 18,19 .
A major challenge for researchers is to obtain a reliable and valid instrument to measure physical activity 20 .For logistical reasons, questionnaires are used in most studies although efforts are being made to use direct measurements more frequently.Therefore, accurate and valid methods for measuring physical activity among adolescents, as well as among other age groups, are needed to determine the dose-response associations between physical activity and health, in order to monitor temporal trends in populations and to make cross-cultural comparisons also determining the effect of interventions 21 .
In Brazil, there are no standardised instruments, direct or indirect, to estimate physical inactivity among young people 22 .All studies have used indirect methods to assess physical activity (i.e., methods that use reports from the individuals themselves).The accuracy of self-reporting is influenced by the respondent's ability to accurately recall all relevant activities, retrospectively.Therefore, indirect methods are subject to recall bias 23 .
Only seven studies used physical activity logs, an instrument in which the individual registers any and all activities performed during the day in short time intervals (typically 15-minute peri-ods) 24 .Most researchers opted to use questionnaires, of which 16 were validated and 14 were unique to the individual study.Although validation studies have low correlation values between the questionnaires evaluated and a gold standard, their use is still recommended over unique, non-validated questionnaires 25 .The use of validated instruments allows for comparisons between different studies.
The use of different methods to assess physical activity could affect the results of the studies.Some authors found higher prevalences when their questionnaires included all domains (commuting to work, leisure-time and sports) and different environments of physical activity (inside or outside school) when compared with other studies that excluded some domains or activities at school.For example, some authors defined physical activity as the practice of sports outside school as well as going to school by bike or walking 26,27 .Other authors considered any kind of vigorous and/or moderate physical activity, inside or outside school, in a structured way or not 28,29,30,31,32,33,34 .The lack of a homogeneous definition of physical activity is an important limitation for pooling, or even comparing, results among different studies.
Another factor that complicates comparisons between the studies of Brazilian adolescents is the lack of standardization for the cut-off point used to classify individuals as "inactive".Most public health authorities agree that adolescents should accumulate at least 60 minutes of moderate-to-vigorous physical activity at least five days a week 7,35,36 , resulting in a cut-off point of 300 minutes per week.However, when addressing cut-off points, studies had noted variability using different values for this classification.
The difficulties in explaining the observed heterogeneity can be attributed to the dispersed nature and small number of studies related to the characteristics analysed.For example, although 16 studies used the same cut-off point for inactivity, 300 min/week, these 16 studies used different questionnaires and considered different areas of physical activity.
Sisson & Katzmarzyk 37 performed a review of studies presenting national prevalences of physical activity (more than 300 min/week).In their review they reported that women were more inactive than men, considering both adults (> 18 years) and adolescents (< 18 years).In studies with young subjects, the prevalence of physical inactivity varied widely among populations.The countries with highest prevalences were Belgium, France, and Tonga.
To the best of our knowledge, there had been no previously published studies using systematic review and meta-analysis tools to investigate the prevalence of physical activity/inactivity among adolescents.

Conclusion
Despite the noted challenges, we believe that this study is the first effort to systematise information regarding physical inactivity, which is largely relevant to the health of adolescents in Brazil.This review highlights the urgent need to use standardised and validated methods in order to measure physical activity.As such, we shall be able to identify specific groups for intervention action, increasing their physical activity and its beneficial effects.
ContributorsL. A. Barufaldi, G. A. Abreu participated in the literature search, article selection, extraction and analysis of the data and in the drafting of the manuscript.E. S. F. Coutinho participated in the analysis of the data and in the drafting of the manuscript.K. V. Bloch participated in article selection, analysis of the data and in the drafting of the manuscript.

Table 1
Characteristics of included studies.

Table 2
Prevalence rates of physical inactivity among Brazilian adolescents -total and by gender.Forest plot of the prevalence rates of physical inactivity among Brazilian adolescents, by region and gender.