Physical Activity, Sedentary Behavior, Cardiorespiratory Fitness and Metabolic Syndrome in Adolescents: Systematic Review and Meta-Analysis of Observational Evidence

Background Metabolic syndrome (MetS) has been diagnosed in adolescents and among the associated factors are low levels of physical activity, sedentary behavior over long periods and low cardiorespiratory fitness. However, specifically in adolescents, studies present conflicting results. The aim of the present study was to conduct a systematic review and meta-analysis of observational studies, in order to map the association between physical activity, sedentary behavior, cardiorespiratory fitness and MetS in adolescents. Methods A search was performed in the databases PubMed, SPORTDiscus, LILACS and the Cochrane Library. For the meta-analysis, the odds ratio (OR) was calculated together with the respective confidence intervals (95% CI), in which the measures of effect were analyzed by dichotomous data (exposure variables) with MetS used as events. Results Eighteen studies were included in the meta-analysis. Primary analysis demonstrated that low levels of physical activity (OR = 1.35 [1.03 to 1.79]; p = 0.03) and low cardiorespiratory fitness (OR = 4.05 [2.09 to 7.87]; p < 0.01) were significantly associated with the development of MetS, while for sedentary behavior, represented by screen time > 2 hours/day, a significant association was not identified (OR = 1.20 [0.91 to 1.59]; p = 0.20). Subgroup analyses demonstrated that the association between low physical activity and MetS was dependent on the use of the accelerometry technique (OR = 2.93 [1.56 to 5.47]; p < 0.01). Screen time > 2 hours/day was significantly associated with MetS only on weekends (OR = 2.05 [1.13 to 3.73]; p = 0.02). With respect to cardiorespiratory fitness, a significant association with MetS was found independent of the maximal oxygen uptake (VO2max) measurement method. Conclusions Low levels of physical activity, low indices of cardiorespiratory fitness and sedentary behavior, represented by screen time > 2 hours/day on weekends, were significantly associated with the development of MetS in adolescence.

Introduction data, a fact which is amplified in young people due to the difficulty teenagers have to accurately remember the intensity, duration and frequency of physical activity [33].
Thus, the objective of this systematic review and meta-analysis was to summarize the results of observational data presented in case-control, cross-sectional or prospective cohort studies that examined the association between physical activity, sedentary behavior, cardiorespiratory fitness and MetS in adolescents from 10-19 years old.

Materials and Methods
This study is characterized as a systematic review, accompanied by meta-analysis, according to the PRISMA protocol (S1 Table. PRISMA Checklist) [34,35] and registered in PROSPERO (CRD42016043113). The inclusion criteria were: a) observational studies (case-control, crosssectional and prospective cohort); b) studies which identified MetS, taking into account criteria adapted for adolescents; c) studies that considered physical activity, sedentary behavior and cardiorespiratory fitness as an exposure variable and MetS as the outcome variable; and d) studies that included adolescents aged 10-19 years.
The exclusion criteria were: a) no observational study design; b) studies with duplicate information already included in other studies; c) studies that did not define MetS, or only considered the risk dimension based on a continuous score; c) MetS not associated with the variables of interest (physical activity, sedentary behavior and cardiorespiratory fitness); d) adult population (> 19 years) or children (<10 years); and e) studies including adolescents with a pathological condition, physical or mental disability.

Databases & search strategy
The search was conducted in the databases PubMed, SPORTDiscus, LILACS and the Cochrane Library without the use of a filter to limit date of publication or language. The final search took place on May 07, 2016. As the search strategy, the following keywords were selected: ("metabolic syndrome" OR "metabolic syndrome x" OR "syndrome x") AND ("physical activity" OR "motor activity" OR "sedentary behavior" OR "sedentary activity" OR "sedentary lifestyle" OR "screen time" OR "physical fitness" OR "cardiorespiratory fitness" OR "exercise" OR "sports") AND ("adolescent" OR "youth" OR "teen" OR "teenager"). The search strategy was adapted for each database when necessary. A manual search taking into account the reference lists of included studies was used to supplement the database search.

Selection of studies
One reviewer (RGO) conducted the initial search strategy in the databases, extracting titles and abstracts. Subsequently, the selection of studies, analysis and data extraction were conducted independently by two authors (RGO and DPG), based on blindly reading the titles and abstracts. Potentially eligible articles were read in full. Subsequently, the disagreements were resolved by consensus between the two authors who used the same form for data extraction.

Data extraction
The following were extracted from each study: a) identification of the authors; b) year of publication; c) country of origin; d) sample size; e) gender and age of the subjects; f) study design (casecontrol, cross-sectional or prospective cohort); g) MetS identification criteria; h) prevalence of MetS; i) indicator used to measure physical activity, sedentary behavior and cardiorespiratory fitness; j) stratification criteria for physical activity, sedentary behavior and cardiorespiratory fitness; k) adjustment variables; and l) principle results.

Evaluation of the methodological quality of the studies
To evaluate the potential risk of bias and methodological quality of the studies, each study was critically analyzed blindly by the authors using an adapted version of the tool proposed by Downs and Black [36]. This tool is comparable to the Cochrane Collaboration tool for assessing risk of bias, and has often been used in systematic reviews and meta-analyses of observational studies [37][38][39][40]. As recommended, the items of the original check list directed to experimental studies and items that did not apply to this study were excluded. The version of the tool used consisted of 10 items wherein, in each item a score was assigned: 1 point (confirmed attribute) or 0 points (unconfirmed attribute). Thus, the studies included in the systematic review and meta-analysis were classified as high (10-9 points), moderate (8-6 points), low (5-4 points) or very low quality (< 4 points). Possible disagreements on the final score were resolved by consensus among the authors.

Definitions
Physical activity was defined as any bodily movement produced by skeletal muscles that resulted in energy expenditure above resting levels. Since physical activity is a complex behavior, it is typically divided into categories, which are mutually exclusive, such as light, moderate or vigorous intensity [41]. On the other hand, sedentary behavior was defined as activities involving low energy expenditure, equivalent to 1.0 to 1.5 metabolic equivalent units (METs), such as remaining seated or lying down, reading, watching TV, playing video games, or other forms of entertainment based on a screen [42]. Furthermore, cardiorespiratory fitness was defined as the ability of the respiratory and circulatory systems to supply fuel during prolonged activities and eliminate fatigue, translated by the maximal oxygen uptake (VO 2 max) [41].
MetS was defined as a cluster of risk factors for cardiovascular disease and type 2 diabetes mellitus, which included: a) high blood pressure; b) increased triglycerides; c) reduced highdensity lipoprotein cholesterol (HDL-c); d) impaired fasting glucose; and e) abdominal obesity [3,4]. The factors were considered independently of the cut-off points established for each of them, which may vary between the different criteria proposed for adolescents [5,6]. To define MetS, different definitions were developed, principally from the criteria adopted by the National Cholesterol Education Program Adult Treatment Panel III (NCEP-ATPIII) [43], with cut-off points adapted to adolescents (eg.: Ford et al. [44], de Ferranti et al. [45], Cook et al. [46], Jolliffe and Janssen [47]), in which any three of the five factors were required to be present. Another definition, created by the International Diabetes Federation (IDF) [48], considers the mandatory presence of abdominal obesity and any other two factors.
Some studies [8,13,15,17,18,20,26] identify MetS using two or more criteria together, which is denominated harmonized criteria [29]. In the present review, MetS was not considered when its identification occurred through the presence of only two risk factors, or when it was considered through a continuous score (Z-score).

Synthesis of results
The descriptive results of the qualitative synthesis of the studies are presented in tables. For the meta-analysis, the measures of effects were analyzed by dichotomous data (low level of physical activity vs. moderate/high level of physical activity, low screen time vs. high screen time, low cardiorespiratory fitness vs. moderate/high cardiorespiratory fitness) and MetS taken as an event in each category. To verify that there is a greater chance for the development of MetS among the different classifications within the exposure variables, the Odds Ratio (OR) was calculated with a confidence interval of 95% (CI 95%). The OR is 1 when there is no association, being significant only when the 95% CI does not pass the value of 1. The Cochrane Q test for heterogeneity was performed, assuming a statistical significance of p 0.10. Heterogeneity was also quantified through the I 2 statistic, where 40% does not indicate significant heterogeneity, 30-60% indicates moderate heterogeneity, 50-90% indicates high heterogeneity and ! 75% indicates considerable heterogeneity [49]. When statistically significant heterogeneities were not identified, the fixed effects model was used; otherwise, we used the random effects models. Statistical significance equivalent to the values of the effect size of the association was considered as p < 0.05. Forest plots were generated for each test ordered by year of publication of the studies. To evaluate the risk of publication bias, a funnel plot was used when there were ! 10 studies in the same meta-analysis. All analyses were performed with the program Review Manager (RevMan) [Computer program], version 5.3, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration.
Data handling. For statistical calculation, all ORs were converted into their natural logarithms (log), accompanied by the respective standard error (SE). Thus, when necessary, it was possible to reverse the order of the categories for a given variable [50]. Whenever available the most adjusted OR (adjusted statistically for different covariates) was taken directly from the publications [14,17,18,20,23,25]. However, when OR scores were not available in the studies included in the systematic review [9,12,15,19,22], absolute data were taken for the total quantity of volunteers and MetS events in each meta-analysis comparison category. For missing data, the authors of the selected studies [8,10,11,13,16,21,[23][24][25][26][27] were contacted on at least three occasions by one of the reviewers (RGO). There was a 64% response rate [8,10,13,16,[23][24][25]. Laurson et al. [8], Fadzlina et al. [10], Múnera et al. [13], Stabelini Neto et al. [16] and McMurray et al. [24] provided the missing data. When the information was not available, estimates were used to obtain the missing data, converting the percentage values (total quantity of volunteers and MetS events in each meta-analysis comparison category) given in the original publication [21,26,27] to absolute values.
Thus, only one study selected in the systematic review [11] was not included in the metaanalysis of physical activity due to lack of data; and two others [23,25] had sufficient information for the meta-analysis on physical activity, but not for sedentary behavior, so could be included only in the first analysis (physical activity). One publication [22] provided the data on the total number of participants in each category of interest, however, the quantity of MetS events did not allow actual calculation of OR. Furthermore, one case-control study [7], was not comparable to the others in the quantitative synthesis (meta-analysis), given its sample selection feature (intentionally included 32 obese young people with MetS). Thus, three studies [7,11,22] were not included in the meta-analysis and could only be considered in the qualitative synthesis.
Primary analysis. In the primary analysis all studies with available data were considered, regardless of methodological quality. For physical activity, studies [9,12,13,16,19,24,26] presenting data on moderate and high levels of practice separately, were brought together in a single group. In relation to the cut-off points for physical activity, the classifications used by each study were considered, which varied according to the measuring instrument used in their designs. For sedentary behavior, which was characterized in all studies by the screen time, a cut-off point 2 hours/day was adopted. This is because in three studies [10,13,14] screen time was considered low when 2 hours/day; in two other studies [12,15], < 2 hours/day; and in a single study [21] 16 hours/week (about 2 hours and 17min/day). For cardiorespiratory fitness, in studies [16,24,27] which presented moderate and high fitness data categories separately, these were brought together in a single group. Regarding the cut-off points, the classifications adopted by each study were considered, primarily distribution tertiles [16,24,27] or low vs. high fitness [8,23].
Analysis of sensitivity and subgroups. Subsequently, in order to verify whether studies with a greater risk of bias could be affecting the results of the meta-analysis, sensitivity analysis was performed, in which studies of low methodological quality were excluded. Several subgroup analyses were also performed, to identify other variables which could influence the results of the primary analysis. Regarding physical activity, the two forms of measurement (questionnaire and accelerometer) were considered separately. For sedentary behavior, the studies that measured the screen time for every day of the week and those that only considered the weekend were considered separately. In the case of cardiorespiratory fitness, the various measurement protocols were considered separately (field tests or laboratory tests, submaximal or maximal). Additional subgroup analyses were also performed involving: a) crude ORs; b) ORs adjusted for confounding variables; c) high prevalence of MetS (> 6.7% [prevalence above the median in relation to the studies included in the meta-analysis]); d) low prevalence of MetS ( 6.7% [prevalence at or below the median in relation to the studies included in the meta-analysis]); e) each diagnostic criteria for MetS.

Qualitative synthesis of the studies
In the databases considered, 773 studies were located. In addition, the manual search enabled the addition of three more studies. After exclusion of duplicate studies in the databases, an initial total of 715 studies was considered. Subsequently, titles and abstracts were analyzed and 628 studies were excluded, leaving 87 studies for analysis of the text in full. After reading the full texts, 66 studies did not meet the eligibility criteria (a complete list of studies excluded at this stage is available in S2 Table). The grounds for exclusion were: study design (6 studies); MetS not linked with outcomes of interest (15 studies); did not consider MetS (20 studies); and age (25 studies). Thus, 21 studies  were included in the qualitative synthesis and of these, 18 [8][9][10][12][13][14][15][16][17][18][19][20][21][23][24][25][26][27] offered sufficient data for inclusion in the meta-analysis (Fig 1).
Regarding the results reported in the selected studies comparing the level of physical activity among adolescent carriers and non-carriers of MetS, four studies found no differences [7,9,13,23]; and two identified less physical activity in the MetS carrier adolescent group [19,24]. In studies in which the prevalence of MetS was compared among adolescents classified into different levels of physical activity, two studies showed no significant differences [15,26], one study found a higher prevalence of MetS according to less physical activity [16] and another study, in contrast, showed a higher prevalence of MetS among adolescents with higher physical activity [14]. Regarding the association between physical activity and MetS, eight studies presented no statistically significant OR values [10][11][12]14,16,17,20,25]. Two studies indicated that adolescents with low levels of physical activity presented a two to five times increase in the chances of having MetS [18,19]; while another study found that high levels of physical activity offered a protective effect related to the appearance of MetS (OR = 0.4 [CI 95%, 0.18 to   [23]. One prospective study found that low levels of physical activity in childhood increased by about five times the odds of the presence of MetS in adolescence [24]. In relation to the adjustments for confounding variables, two studies ignored this procedure [12,20]; and four studies demonstrated adjusted OR values, but did not report the variables selected in the adjustments [10,14,17,18]. The other studies performed adjustments for age and/or gender [11,16,19,[23][24][25]. Only one study adjusted for body mass index [24] economic situation [19], sedentary behavior and smoking [25].

Sedentary behavior
For the nine studies that associated outcomes involving sedentary behavior and MetS (Table 2 [7,10,12-15,21,23,25]), the year of publication ranged from 2008 to 2016. Of these, four studies were from Asia [10,12,14,21], three from South America [7,13,15], one from Europe [23] and one from North America [25]. The total number of participants was equal to 8,680 (ranging from 96 [7] to 2,900 adolescents [23]). One study included only overweight and obese adolescents [13]. Only one case-control study was included [7] and no cohort studies. Other studies presented a cross-sectional design [10,[12][13][14][15]21,23,25]. MetS was primarily diagnosed by IDF criteria (3 studies [10,12,23]). Two studies used harmonized criteria [13,15], while the remainder used different criteria adapted for adolescents, from the NCEP-ATPIII [7,14,21,25]. The prevalence of MetS ranged from 0.8% [23] to 13% [12,14] (except for the case-control study [7], in which a sample of 32 adolescents with MetS had been previously selected). Details on the occurrence of MetS events in different classifications of sedentary behavior are presented in S5 Table. Sedentary behavior was treated by screen time in all studies, primarily identified by questionnaire [10,12,14,15,21,23,25]. Two studies used recall [7,13]. Three studies investigated only TV time [7,13,23]; two studies TV time and recreational computer use [14,21]; and the remaining four studies reported any form of screen-based entertainment [10,12,15,25]. Regarding the results reported in the literature, of the studies that identified differences in screen time among adolescent carriers and non-carriers of MetS, two showed no significant differences [7,13]; and two other studies found that screen time was higher in the group with MetS [21,23]. In studies that established comparisons between the prevalence of MetS and different screen times, one study identified no significant differences (< 2 hours/day vs. ! 2 hours/day) [15]; another study found a higher prevalence of MetS according to increased screen time, identified by distribution quartiles [21]; and a third study found a higher prevalence of MetS among adolescents who reported more TV time, however, this was not the case for recreational computer use ( 2 hours/day vs. > 2 hours/day) [14]. In another study, the prevalence of MetS increased with higher screen time ( 1, 2, 3, 4 ! 5 hours/day). Regarding the association between MetS and screen time, two studies found no statistical significance [10,12]. One study observed that higher TV time at weekends (> 2 hours/day) could double the chances of developing MetS [14]. Another study found that adolescents with screen time ! 35 hours/week tended to be twice as likely to develop MetS compared to their peers with screen time 16 hours/week [21]. Another study identified that screen time ! 5 hours/ day increases the odds of developing MetS by about three times compared with screen time 1 hour/day [25]. Regarding adjustments for confounding factors, two studies demonstrated an adjusted OR, but did not report the variables selected for adjustment [10,13]. Other studies established adjustments mainly for age [12,21,25]. Only one study adjusted for smoking and physical activity [25], or gender, economic status and area of residence [21].

Cardiorespiratory fitness
The six studies that associated outcomes involving cardiorespiratory fitness and MetS (Table 3 [8, 16,[22][23][24]27]) were published between 2007 and 2015. Of these, three studies were from Europe [8,22,23], two were conducted in North America [24,27] and one in South America [16]. In total, they included 5,748 participants, ranging from 379 [8] to 2,446 adolescents [23]. No case-control studies were found in the literature and only one prospective cohort study [24]; the other studies had cross-sectional designs [8,16,22,23,27]. MetS was diagnosed in two studies by means of the criteria proposed by the IDF [22,23] and one study by harmonized criteria [8], the other three studies used criteria derived from the NCEP-ATPIII, adapted for teenagers [16,24,27]. The prevalence of MetS ranged from 0.8% [23] to 7.7% [16]. Details on the occurrence of MetS events in different classifications of cardiorespiratory fitness are presented in S6 Table. To establish estimates of cardiorespiratory fitness, two studies used similar field tests (20-meter shuttle run test and endurance run test-PACER) [16,22], while the other four studies made use of a treadmill or cycle ergometer through maximal [8,23] and submaximal load tests [24,27].
Regarding the results reported in publications, the studies comparing scores equivalent to VO 2 max, among adolescent carriers and non-carriers of MetS, identified lower cardiorespiratory fitness in adolescents with MetS [23,24]. Studies comparing the prevalence of MetS between different levels of cardiorespiratory fitness found higher prevalence, according to lower levels of cardiorespiratory fitness [16,22]. For measures of association between cardiorespiratory fitness and MetS, two studies demonstrated that higher cardiorespiratory fitness may offer a significant protective effect for MetS (OR = 0.43 [CI 95%, 0.18 to 0.88] and OR = 0.01 [CI 95%, 0.00 to 0.07]) [23,27]. Similarly, two other studies demonstrated that low cardiorespiratory fitness increased the risk of the onset and development of MetS by three to five times [8,16]. The one prospective cohort study pointed out that low cardiorespiratory fitness in childhood may increase by six times the chances of identifying MetS in adolescence [24]. Regarding adjustments for confounding variables, one study chose not to perform this procedure [8], while the remaining studies adjusted primarily for age and/or gender [16,23,24,27]. Furthermore, one study adjusted for body mass index [24] and another for ethnicity, smoking, economic status, and intake of lipids and carbohydrates [27].
For the sensitivity analysis, studies were excluded that addressed the association between MetS and physical activity [10,17] or sedentary behavior [10], due to low methodological quality. In the case of cardiorespiratory fitness, even considering that no study was categorized as low methodological quality, one study was excluded [8] due to possible increased risk of bias, for presenting lower scores equivalent to methodological quality. However, results found in the primary analysis showed no alterations after exclusion of these studies (S1, S2 and S3 Figs).
Subgroup analyses. For physical activity, two subgroup analyses were performed (Fig 2), based on the different measuring instruments used in the studies. When considering the studies that used questionnaires (OR = 1.23 [CI 95%, 0.93 to 1.63] p = 0.15, n = 15,194, studies = 13, I 2 = 61%), no significant association was observed between physical activity and MetS. However, when measured through the use of accelerometers, a low level of physical activity was significantly associated with a higher chance of developing MetS (OR = 2.93 [CI 95%, 1.56 to 5.47] p = 0.0008, n = 2,030, studies = 2, I 2 = 0%).
Regarding sedentary behavior, two subgroup analyses were carried out (Fig 3). In the first, studies were considered that measured screen time all week. In this case, there was no significant association between screen time and MetS (OR = 1.03 [CI 95%, 0.75 to 1.42] p = 0.84, n = 2,261, studies = 4, I 2 = 24%). However, when taking into account screen time specifically at the weekend there was a significant association (OR = 2.05 [CI 95%, 1.13 to 3.73] p = 0.02, n = 1,620, studies = 2, I 2 = 0%), between higher screen time and an increased chance of developing MetS.
For cardiorespiratory fitness, analyses were grouped into three types of tests to estimate VO 2 max (Fig 4). When considering studies that used maximal field tests (20-    Additional analyses of subgroups were performed for: crude ORs (S4, S5 and S6 Figs) vs. Adjusted ORs (S7, S8 and S9 Figs), studies with a high prevalence of MetS vs. studies with a low prevalence (S10, S11 and S12 Figs), and according to the diagnostic criteria used (S13, S14 and S15 Figs). When crude OR data were selected (without adjustments for confounding variables), physical activity was not significantly associated with MetS, either through the use of a questionnaire or the accelerometry technique (S4 Fig). High sedentary behavior  When only studies that found a high MetS prevalence were grouped, no significant association was found with physical activity, however, for studies that found a low MetS prevalence, a significant association was observed between low physical activity level and a higher risk of MetS (OR = 1.78 [CI 95%, 1.27 to 2.50] p = 0.0009, n = 9,128, studies = 8, I 2 = 22%, S10 Fig).
For the subgroup analysis in which each MetS diagnostic criterion was verified separately, a low level of physical activity was associated with a greater risk of developing MetS through the Jolliffe and Janssen [47] diagnostic criteria (OR = 2.56 [CI 95%, 1.29 to 5.08] p = 0.007, n = 2,192, studies = 2, I 2 = 0%), or when using harmonized criteria (OR = 1.36 [CI 95%, 1.01 to 1.84] p = 0.04, n = 9,712, studies = 6, I 2 = 49%). For the remaining diagnostic criteria, there were no significant associations between physical activity and MetS (S13 Fig). Regarding sedentary behavior, no significant association was observed when the studies were grouped by diagnostic criteria (S14 Fig). However, low cardiorespiratory fitness was significantly associated with the development of MetS, independent of the diagnostic criteria used (S15 Fig).

Agreements and disagreements with other studies
To our knowledge, to date, no meta-analysis involving adolescents has investigated the variables selected in this study. A previously performed systematic review suggested evidence that low levels of physical activity may be associated with an increased risk of the onset and development of MetS in adolescents [31]. Primary and sensitivity analysis performed in the present study confirmed the existence of a significant association between physical activity and MetS. From the qualitative analyses of the studies included in the systematic review, it was not possible to predict this finding, since only four [18,19,23,24] of the 12 selected studies that sought to identify an association between physical activity and MetS found significant results.
Among the selected studies, only two used the accelerometry technique to measure physical activity [19,23], which provides more reliable results [33]. Nguyen et al. [19] found that being located in the lowest quartile of physical activity increased the chances of developing MetS by more than five times, and Ekelund et al. [23] identified a lower risk (OR = 0.4 [CI 95%, 0.18 to 0.88]) for the development of MetS in adolescents with higher levels of physical activity. These findings available in individual studies were confirmed in the subgroup analysis performed in the present study, in which a significant association between MetS and physical activity was not identified when evaluated by self reported measures (questionnaire), however, in the analysis that considered the two studies in which physical activity was scaled using accelerometers, it was found that low levels of physical activity increase the odds of developing MetS by approximately three times.
In adolescents the use of self reported measures of physical activity can increase the possibility of error, since this population tends to present more difficulty remembering and reporting accurately the intensity, frequency and duration of activities [33]. Previous studies indicate that self reported measures tend to overestimate the actual level of physical activity [51]. In addition, some questionnaires have not been sufficiently validated against objective methods, creating doubts about their effective ability to properly evaluate the level of physical activity, especially among young people [33].
Thus, it is possible to assume that the effect size of the association measure of the primary analysis for physical activity was underestimated. In principle, it was found that low levels of physical activity increase the chances of developing MetS by 35%. However, the subgroup analysis with measurement of physical activity using accelerometers demonstrated about a three times higher chance of developing MetS, suggesting that this could be the actual size of the effect if other studies including more precise measurements of physical activity had been available. However, the low number of studies in this subgroup analysis restricted safer extrapolations. It should also be considered that a significant association between physical activity and MetS observed using the accelerometry technique was dependent on OR adjusted for confounding variables (age, gender, and nationality), since when only crude ORs were considered, no significant association was observed.
The subgroup analyses also showed that a low level of physical activity was significantly associated with a 78% greater risk of developing MetS when grouping studies that found a low prevalence of the syndrome, whereas a significant association was not found in studies that observed a high prevalence. This may have been due to the fact that studies that find a high prevalence of MetS usually rely on less rigorous criteria to diagnose the syndrome. It has been shown that less rigorous criteria (e.g.: de Ferranti et al. [45]) can overestimate the proportion of adolescents with MetS [6]. This argument is reinforced by the subgroup analyses in which the studies were separated by diagnostic criteria. In this case, the association between low levels of physical activity and MetS was maintained only for the Jolliffe and Janssen criteria [47] and for studies using harmonized criteria, in which a low prevalence of MetS was found in the majority of studies [13,15,24,25,26]. The exception was the IDF criterion, which despite being rigorous to classify adolescents with MetS [6] presented no significant association in the subgroup analysis.
In relation to sedentary behavior, an important limitation was the use of self-reported measurements for screen time in all studies, possibly sacrificing the quality of the treated measure. Another fact that should be considered is the cut-off point adopted for the screen time in several studies [10,[12][13][14]15], which typically referred to 2 hours/day. This cut-off point was defined by consensus among specialists who sought to establish guidelines based on evidence for health with respect to sedentary behavior in children and adolescents [52].
However, in the primary and sensitivity analyses performed based on this cut-off point, there were no significant associations between screen time and MetS. Individual studies included in the systematic review that showed significant associations used different cut-off points. Kang et al. [21] found that screen time equivalent to ! 35 hours/week (approximately ! 5 hours/day) compared to 16 hours/week (approximately 2 hours and 17 min/day) lead to two times the chances of developing MetS. Similarly, Mark and Janssen [25] only identified an association when considering screen time ! 5 hours/day versus 1 hour/day, indicating in this case close to three times more risk of developing MetS. The exception was the study of You and Son [14], who found twice the risk for MetS when screen time was > 2 hours/day, however, specifically on weekend days.
The subgroup analysis indicated that screen time > 2 hours/day on the weekend, revealed a two times greater chance of developing MetS in adolescents, while an association involving days of the week, showed no statistical significance. Kang et al. [21], when analyzing the association between screen time and MetS, considering only the weekend, identified a risk almost three times greater for MetS, but only for the screen time strata ! 7 hours/day versus 3 hours/day. In the same sample, the researchers found no association when considering only days of the week (excluding weekends). In particular, there seems to be evidence in the sense that there may be a greater association between screen time at weekends and MetS in adolescents, than during weekdays. However, the subgroup analysis showed that when using only OR adjusted for confounding variables, the significant association between screen time at weekends and MetS disappeared.
When only studies with a high or low prevalence of MetS were considered separately, no significant association was observed between screen time and MetS, in the same way as when studies were grouped by diagnostic criteria. In this case, the same limitations as previously presented should be considered (use of self-reported measures, cut-off point 2 hours/day, and low number of studies in the meta-analysis). In addition, the primary analysis had already indicated that there was no significant association between sedentary behavior and MetS.
When establishing associations between cardiorespiratory fitness and MetS in adolescents, the primary analysis indicated that lower VO 2 max values can present an approximately four times greater chance of MetS. Data analyzed qualitatively in the systematic review had already indicated this trend. The five selected individual studies [8,16,23,24,27] showed significant associations between cardiorespiratory fitness and MetS in adolescents.
The subgroup analyses demonstrated that the significant association is independent of the three different types of tests used. Despite having provided a smaller effect size (OR = 2.05 [CI 95%, 1.02 to 4.09]), the test used (20-meter shuttle run test) by Stabelini Neto [16], has the advantage of being administered in large numbers of subjects simultaneously, in a short time and at low cost. In addition, a previously performed meta-analysis demonstrated that VO 2 max estimated using the 20-meter shuttle run test presents good correlation (r = 0.78) with methods involving gas exchange in adolescents [53].
Given that maximal and submaximal tests involving cycle ergometers and treadmills may not be feasible in many places, the multi-stage 20-meter test constitutes a useful alternative to estimate cardiorespiratory fitness, especially for adolescents in a school environment [53]. This type of test can be used in an attempt to identify adolescents with low cardiorespiratory fitness and a consequently greater metabolic risk. Moreira et al. [22], using a similar test (PACER), found a significantly higher prevalence of MetS in the group of adolescents with low cardiorespiratory fitness (8.5%) compared with the group of adolescents with high cardiorespiratory fitness (0%). However, it was only possible to select one study involving this type of testing for the subgroup analysis, which could limit any generalizations.
Finally, the association between low cardiorespiratory fitness and MetS occurred regardless of the use of crude OR or adjusted for confounding variables. In addition, considering the strong association between these variables already presented in the primary analysis, it was observed that a significant association between low cardiorespiratory fitness and MetS remained independent of the prevalence found, or of the diagnostic criteria used.

Quality of evidence
Of the 15 studies included in the meta-analysis involving physical activity, there was representation from five different geographical regions (Asia [9,10,12,14,19,20] Latin America [13,15,16], North America [24,25,26], Europe [17,23] and Africa [18]), while for sedentary behavior only two regions (Asia [10,12,14,21] and Latin America [13,15]) and for cardiorespiratory fitness, three regions (North America [24,27], Europe [8,23] and Latin America [16]). Thus, it is not possible to extrapolate the results to all locations, since there was no reasonable representation from the different regions. In contrast, the quality of the studies included in the systematic review was considered moderate, with a mean score of 7.8 ± 1.5 points. Only two studies [10,17] presented low quality indicators and sensitivity analysis demonstrated that the exclusion of both studies did not change the results of the primary analysis.
However, of the 18 studies included in the meta-analysis, only four [12,14,16,27] presented data separately for boys and girls, which made subgroup analysis impossible for this characteristic. Similarly, the studies did not present data for different ages in adolescence, preventing subgroup analyses for this factor. Furthermore, MetS was identified according to five different diagnostic criteria (IDF [10,12,19,22,23] Cook [11,14,16], Jolliffe and Janssen [24,25,27], Ford [7,21], and de Ferranti [9]), as well as some researchers [8,13,15,17,18,20,26] using harmonized criteria. This is configured as a limiting factor to be taken into account when interpreting the results, since the prevalence of MetS in adolescence can vary significantly when diagnosed by different criteria [6]. In addition, some studies did not adjust the measures of association taking into consideration possible confounding variables [8,12,20], while others, despite reporting the use of covariables, did not specify which ones were used [10,14,17,18].

Potential biases in the review process
In the systematic review and meta-analysis only observational studies were included, which increases the risk of bias, especially since the majority of the studies selected were cross-sectional, thus hindering the performance of causal inferences. The number of studies available for meta-analysis should also be raised as another limitation. Only one analysis was carried out with ! 10 studies (primary analysis, the odds ratio for metabolic syndrome in adolescents with moderate/high levels of physical activity versus low levels of physical activity), thus allowing visualization on the funnel plot (S16 Fig), which showed no evidence of publication bias. Moreover, the majority of analyses used random effects models, considering the heterogeneity of the data, which reduces the effect of the associations and, therefore, the results should be interpreted with caution. Finally, the search did not extend to all existing databases. Nonetheless, we performed searches in three primary databases (PubMed, SPORTDiscus, and LILACS) and one secondary database (The Cochrane Library). In addition, we conducted a thorough search of all the references of the studies included in the review in an attempt to find further studies.

Implications for research
The study demonstrated that low levels of physical activity were significantly associated with the development of MetS in adolescents; however, the subgroup analyses indicated that a significant effect was dependent on the measuring instrument (accelerometer). The analysis involving the self reported measurements did not indicate significant association, suggesting that the effect size of the primary analyses may have been underestimated. Thus, due to the limitations of questionnaire use in adolescents, future studies should prioritize the use of objective measures, aiming to investigate the association between physical activity and MetS in this population. Other issues that should be considered in future studies are the importance of using adjustments for covariables, in addition to preferentially using more rigorous criteria to diagnose MetS.
As for sedentary behavior, there was no significant association between screen time > 2 hours/day and MetS in adolescents. Although the subgroup analysis indicated a significant association when taking into account only screen time over the weekend, this association disappeared when OR adjusted for covariables was used. In view of this, it is suggested that future studies investigate screen time separately for weekdays, weekends and throughout the week. They should also be encouraged to perform analysis using different cut-off points for screen time, not solely > 2 hours/day. In addition, whenever possible, objective measures should be employed, avoiding the limitations related to self reported measures.
Finally, in relation to cardiorespiratory fitness, analyses showed the existence of a significant association between low VO 2 max and MetS. However, the primary analysis included few studies, preventing visual inspection of the funnel plot, as a way to detect the existence of possible unpublished studies due to no significant results. In view of this, in an attempt to verify whether the findings of this systematic review and meta-analysis are sustained, it is suggested that further studies be carried out, in order to identify the association between cardiorespiratory fitness and MetS in adolescent populations.

Implications for practice
Among the variables analyzed in this systematic review and meta-analysis, a low level of physical activity, screen time > 2 hours/day at weekends and low cardiorespiratory fitness were the indicators that demonstrated significant associations with MetS in adolescents. Thus, an increase in time spent in physical activity of moderate to high intensity and reduced screen time of 2 hours/day at weekends should be encouraged, especially in adolescents with low cardiorespiratory fitness who are exposed to increased metabolic risk.