Searches

A systematic literature search of EBSCOhost (Academic Search Complete, Education Source, ERIC, Global Health, MEDLINE Complete, SPORTDiscus) and Scopus databases was conducted in October 2017. The PICO-based search strategies for the databases included the following key words in three main areas: 1) Population “children and adolescents” (e.g., youth); AND 2) Intervention/exposure “activity” (e.g., physical activity, sedentary behaviour) AND “activity patterns” (e.g., bouts, transitions, breaks) AND “objective measurement” (e.g., accelerometer, inclinometer); AND 3) Outcome “cardio-metabolic health” (e.g., BMI, blood pressure). The EBSCOhost search strategy is shown in S1 Table as an example. Articles were extracted and imported into EndNote X7.5 (Thomson Reuters, New York, USA).

Study inclusion and exclusion criteria

Studies were eligible for inclusion in this review if they: a) used a quantitative research design (e.g., cross-sectional, randomised controlled trial); b) included participants aged 5–19 years (i.e. likely to be attending primary or secondary school); c) reported associations between activity patterns of any intensity (i.e., sedentary, light, moderate, vigorous, very hard) with cardio-metabolic risk factors (i.e., adiposity, blood lipids, inflammatory biomarkers, endothelial function biomarkers, blood glucose, vascular health, fitness, or summary cardio-metabolic scores); d) objectively measured activity patterns (e.g., accelerometer); e) were published between January 1980 and October 2017; and f) were published in peer-reviewed journals in English or Dutch. Studies that were an advanced publication ahead of print and had a unique digital object identifier (DOI) were eligible for inclusion. To form a dataset representative of the targeted population (i.e., children and adolescents 5–19 years old), studies were excluded if they specifically focused on special/clinical populations (e.g., elite athletes, type 1 diabetes, being treated for obesity/overweight, children and adolescents chosen for family history of specific diseases).

Data extraction strategy

Titles and abstracts were screened by three independent reviewers (LAB, SEC and SJJMV), and full text copies were obtained for all articles that met the initial screening criteria or where the appropriateness of the article was unclear. Three reviewers (NDR, SEC and SJJMV) then independently reviewed the full text of these papers for eligibility for inclusion. Inconsistencies between reviewers were discussed and resolved, and if a consensus could not be reached, the papers were discussed by all authors. Reference lists of included articles were also reviewed to identify any additional studies. Data were extracted by one reviewer (SJJMV) and checked and verified for 15% of articles by another reviewer (LAB) for consistency. Results were consistent and both reviewers agreed on the direction of associations. For the remaining articles, if the reviewer (SJJMV) was uncertain about the interpretation of included results, these were discussed amongst the team members. If more clarification was needed, the authors were contacted via email for clarification. Data extraction was undertaken using a standardised form which included all relevant PICO components; Population (e.g., sample size, sex, BMI, age, and school grade), Intervention/exposure (i.e., activity pattern), and Outcome (i.e., cardio-metabolic risk factors). Information on study characteristics (e.g., study design, type of objective measurement) and statistical methods (e.g., correlations, regression) was also extracted.

Risk of bias assessment

Information on risk of bias (ROB) for individual articles was extracted by two reviewers (LAB and SJJMV) using a tool based on items of the ‘EPHPP Quality Assessment Tool for Quantitative Studies’ [31] (see S1 Text). Four methodological components were assessed in eight items: 1) Selection bias (e.g., sample representativeness); 2) Confounders (e.g., control for relevant confounders); 3) Data collection methods (e.g., whether the methods were reported to be valid and reliable); and 4) Withdrawals and drop-outs (e.g., number of withdrawals) [31]. Each component was given a quality rating of weak, moderate or strong, based on the accompanying instructions for the tool [31]. Studies that received no weak components on the four ROB components were classified as ‘low ROB’. Studies that received one weak rating were classified as ‘medium ROB’, while studies that received two or more weak ratings were classified as ‘high ROB’ [31]. Results were consistent between the reviewers and any disagreements were discussed until a decision was made.

Data synthesis

Any type of activity pattern (e.g. sporadic and prolonged bouts, breaks) of all intensities (i.e., ranging from sedentary to very hard physical activity [VHPA]) was included if it was examined in relation to cardio-metabolic risk factors. According to STROBE (i.e., STrengthening the Reporting of OBservational studies in Epidemiology) guidelines [32], both unadjusted and fully-adjusted models should be presented in epidemiological studies. However, most studies [33–49] (17/29; 59%) included in this review only presented adjusted models (including adjustments for BMI, total MVPA, and age, for example), and six studies (21%) [50–55] presented only unadjusted models. The remaining six studies (21%) [14, 56–60] presented both. Fourteen of the 29 included studies (48%) adjusted for either total MVPA [14, 33, 36, 38, 40–42, 58], sedentary time [60], or both MVPA and sedentary time [34, 37, 45, 48, 59]. The remaining 15 studies examined activity patterns without adjustment for total volume of sedentary time and/or physical activity [35, 39, 43, 44, 46, 47, 49–57]. Consequently, the decision was made to synthesize fully-adjusted models where possible, otherwise unadjusted models were reported. This maximised the opportunity to examine associations between activity patterns with cardio-metabolic health, regardless of the total volume of physical activity and sedentary time.

There was substantial heterogeneity in the definition of activity patterns, variables adjusted for, and effect estimates of cardio-metabolic risk factors (see S2 and S3 Tables); therefore, a meta-analysis was considered inappropriate for this review [61]. Studies reporting observational associations (i.e. cross-sectional and longitudinal studies) [14, 33, 34, 36–47, 49, 51, 53, 55–60] were reported separately from those examining acute effects (i.e. experimental studies) [35, 48, 50, 52, 54]. The results from observational studies are shown in Table 1 and S4–S9 Tables, with each reference number representing one association examined in the corresponding article. Studies reported significantly beneficial (B), significantly detrimental (D), or no significant evidence of an association (NS) between a specific activity pattern (exposure; e.g., >5-min bout of MVPA) and a cardio-metabolic health risk factor (outcome; e.g., BMI). In order to systematically synthesise the observational findings, previously used strategies were considered [62–65]. Consistent with previous systematic reviews [18, 63, 66, 67], only ‘frequently examined associations’ [63] (i.e., if the specific association was investigated at least four times) were discussed. The bold numbers in the right hand columns of Table 1 and S4–S9 Tables represent that specific activity patterns which were examined at least four times. As some studies reported multiple subgroups, only the frequently (≥4 times) investigated associations across multiple studies (i.e., as opposed to within one study) were discussed in the corresponding paragraphs [63]. Given the diversity of outcomes reported, specific outcomes (e.g. body fat, BMI) were grouped under broader outcomes (e.g., adiposity) and discussed. All reported significance levels for the associations were set at p<0.05.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Studies reporting beneficial, non-significant and detrimental associations of activity patterns with adiposity risk factors.

https://doi.org/10.1371/journal.pone.0201947.t001

Review statistics

Extracted data were analysed in October-November 2017. A flowchart of the systematic literature search following PRISMA guidelines [29] is presented in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow chart of the systematic literature search.

From: PRISMA Group [29].

https://doi.org/10.1371/journal.pone.0201947.g001

Detailed characteristics of the 29 included articles are presented in S3 Table. Sixteen studies were conducted in North America [35–39, 41, 44, 46–49, 51, 52, 57–59], eight in Europe [33, 34, 40, 43, 45, 55, 56, 60], and five in Oceania [14, 42, 50, 53, 54]. Sample sizes ranged from nine to 2754 participants. Thirteen studies focused on children aged 5-≤12 years only [14, 35, 37, 39, 44, 47, 51–53, 55, 56, 58, 59], six papers focused only on adolescents aged ≥12–18 years [40–43, 49, 50], and ten papers focused on both age groups [33, 34, 36, 38, 45, 46, 48, 54, 57, 60]. Four papers used data drawn from the same sample (U.S. National Health and Nutrition Examination Survey; NHANES 2003–06) [36, 41, 44, 46]. All studies were published from 2009 onwards, with 13/29 (45%) [14, 33–35, 41–43, 45, 50, 52, 54, 58, 60] papers published from 2015 onwards.

S3 Table shows all significant findings reported in the included studies. Ten studies examined physical activity patterns (including VHPA, Vigorous Physical Activity [VPA], MVPA, Moderate Physical Activity [MPA] and Light Physical Activity [LPA]) [39, 44, 46, 47, 49, 51, 55, 56, 58], while 19 studies examined sedentary patterns [14, 33–38, 40–42, 45, 48, 50, 52–54, 57, 59, 60]. Though some studies acknowledged other intensities by adjusting for total volumes of physical activity or sedentary time, only one study specifically examined patterns across the activity spectrum (i.e., low and high intensities) [43]. One study adjusted for total volume of LPA, when examining sedentary bouts and breaks [59]. Detailed information on the adjustments used in each study, including total volumes, are shown in S3 Table.

The patterns assessed mainly consisted of the number of physical activity/sedentary bouts/breaks of a certain duration (e.g., the number of ≥5-min MVPA bouts), the total time spent in physical activity/sedentary bouts of a certain duration (e.g., minutes spent in ≥5-min MVPA bouts), or the intensity of activity/sedentary bouts (e.g., how high was the intensity of ≥5-min bouts). The fragmentation of physical activity and sedentary behaviour and distinct classes of physical activity patterns were also assessed, albeit less frequently. In observational studies, definitions of activity patterns varied substantially in terms of bout lengths (e.g., ≥4-s to ≥20-min bouts for physical activity; ≥1-min to ≥2-h bouts for sedentary behaviour). Of the studies investigating sedentary patterns, ten included breaks (i.e., a non-sedentary bout in between two sedentary bouts [3]) in their assessment of sedentary patterns [14, 34, 36–38, 40, 53, 57, 59, 60]. In the experimental studies [35, 48, 50, 52, 54], all used cross-over designs that included at least two conditions, of which one was uninterrupted sitting. The uninterrupted sitting conditions ranged from three to eight hours. The applied intensities for active interruptions were based on percentages of the ventilatory threshold [35, 52] and VO₂-peak [48], as well as with preselected activity intensities (measured using cut-points; [54]) or standardised exercises [50].

Thirty-seven different cardio-metabolic risk factors were included in the review and are presented in S2 and S3 Tables. For the purpose of this review, these cardio-metabolic risk factors were classified into the following health outcomes; adiposity, blood lipids, inflammatory and endothelial function biomarkers, glucose metabolism markers, vascular health, fitness, cardio-metabolic summary scores and cortisol.

Risk of bias assessment

Three (10%) studies had the highest possible rating (i.e., low ROB) on the selection bias components, 11 (38%) adjusted for the desired confounders, ten (34%) used valid measures of activity patterns and cardio-metabolic risk factors, and 13 (45%) had minimal withdrawals and dropouts. Consequently, five articles (17%) were classified as ‘low ROB’ [14, 37, 40, 45, 60], 15 (52%) as ‘medium ROB’ [33–36, 38, 39, 41–44, 46, 48, 51, 53, 56], and nine (31%) as ‘high ROB’ [47, 49, 50, 52, 54, 55, 57–59]. Total ROB scores for each study are presented in S3 Table.

Observational associations between activity patterns and cardio-metabolic risk factors

Experimental studies

Five experimental studies were included in this review [35, 48, 50, 52, 54]. The acute effects of (un-)interrupted prolonged sitting were investigated by applying intensities based on the ventilatory threshold [35, 52] and VO₂-peak [48], or activity intensity cut-points [54] and preselected standardised exercises [50] for their sit and walk/exercise protocols. The uninterrupted sitting conditions were of three [35, 52], six [50, 54] and eight [48] hours duration.

Two experimental studies used a sitting protocol and investigated children aged between and 7–11, however, found contrasting results [35, 52]. Belcher and colleagues [35] tested whether three hours of interrupting sitting with moderate-intensity walking bouts every 30 min improved glucose tolerance and found that interrupted sitting resulted in significantly lower insulin, C-peptide and glucose [35]. McManus and colleagues utilised a protocol with moderate-intensity cycling bouts to interrupt sitting every hour. No significant differences between conditions were observed [52].

Fletcher and colleagues [50] focused on adolescents and compared uninterrupted sitting with 2-min activity breaks involving LPA body-weight resistance activities every 18 min. Compared to uninterrupted sitting, the breaks condition elicited a lower postprandial glucose area under the curve response after the first and second standard-energy meal, however, not for the entire trial period or for total area under the curve [50]. Similarly, Ross and colleagues [54] found no differences in triglyceride concentration between six hours uninterrupted sitting and sitting interrupted by short bouts of moderate intensity exercise every 30 min in children and adolescents. Saunders and colleagues [48] examined compared eight hours of uninterrupted sitting with eight hours of sitting interrupted by light intensity walking breaks every 20 min or structured physical activity and found no significant differences in insulin, glucose, triglyceride, HDL cholesterol, and LDL cholesterol in children and adolescents [48].

While similar protocols have been used across experimental studies included in this review [35, 48, 50, 52, 54], the risk factors assessed have been inconsistent. Some evidence suggests that breaking up sitting may influence glucose [35, 50]; few consistent significant results have been found in relation to specific patterns of activity.

Risk of Bias and associations between activity patterns and cardio-metabolic risk factors

Fifteen studies found evidence of a significant beneficial association between an activity pattern and a cardio-metabolic risk factor [35, 38–40, 43, 44, 46, 49, 50, 55–60]. In contrast, seven studies found evidence of a significant detrimental association between an activity pattern and a cardio-metabolic risk factor [33, 34, 37, 38, 43, 45, 59]. Of these studies (i.e., the 19 studies finding any significant results), six (32%) were of high ROB [49, 50, 55, 57–59], nine (47%) were of medium ROB [33–35, 38, 39, 43, 44, 46, 56], and four (21%) were of low ROB [37, 40, 45, 60]. The remaining ten studies did not finding any significant results. Of these studies, three (30%) were of high ROB [47, 52, 54], six (60%) were of medium ROB [36, 41, 42, 48, 51, 53] and one (10%) was of low ROB [14]. ROB scores for each study are presented in S3 Table.

Strengths and limitations

This review is the first to examine associations between specific activity patterns across the entire activity spectrum and a wide range of cardio-metabolic risk factors in children and adolescents. To date, most reviews that have examined activity patterns have not considered which specific activity patterns of activity may be important for health as their primary focus was on total volumes of physical activity and/or sedentary behaviour. Typically, these reviews do not distinguish between different patterns (e.g. frequency, duration, and intensity of bouts) (e.g., [15, 18]), despite this information having the potential to inform future interventions and public health guidance on how to accumulate physical activity and sedentary behaviour (e.g., bouts of particular durations) to benefit health.

There are several limitations which should be taken into account, such as the use of summative evidence coding approach [63] and that some papers (e.g., [38]) included multiple subgroups and/or time points and consequently dominated the evidence database. This may have led to erroneous interpretation of potential beneficial or detrimental associations. However, summative evidence coding [63] is a commonly used approach within this field of research, and did allow more systematic comparisons across studies than a narrative review. The reason for not undertaking a narrative synthesis is that such approaches typically describe and discuss the state of the literature, providing a narrative interpretation and critique of a broader area [68]. Given the number of studies included, this review focused on summarising data [68] from the different studies to highlight how activity patterns are associated with health, and to identify existing gaps. Ultimately, this review highlights the variability defining activity patterns and limitations in evidence, and is the first step towards a more standardised assessment. Nevertheless, following data reporting, results were checked to explore if the results would have differed if we would had only reported one association per study (regardless of the total number of associations reported within a study), but this did not make any difference to the conclusions.

There are also a number of methodological issues to consider that may have influenced the ability to detect associations. Firstly, there were differences in the confounders across studies. Some studies adjusted for total physical activity and/or sedentary time and some did not. Fully-adjusted models were included to examine activity patterns, regardless of total time spent in different intensities. In addition, the chosen intensity thresholds and corresponding cut-points, and epoch lengths (i.e., 2 s to 60 s) varied amongst included studies. Children and adolescents have sporadic activity patterns, with the majority of high intensity activities lasting less than 10 s [24, 69]. Shorter epoch lengths have been shown to minimise errors in measuring these sporadic behaviours in children [70, 71]. Overall, the variation in confounder adjustment, chosen intensity thresholds and corresponding cut-points, and epoch lengths, may have contributed to the lack of consistent associations between activity patterns and risk outcomes observed.

Despite the aforementioned issues, this review identified a number of gaps in current knowledge. Firstly, most studies (21/29; 72%) focused on markers of adiposity (e.g., BMI, waist circumference). Fewer studies have investigated associations between activity patterns and blood biomarkers such as lipids (28%), inflammatory and endothelial biomarkers (17%), or glucose metabolism biomarkers (24%). This is consistent with a review focusing on relationships of combinations of physical activity, sedentary behaviour and sleep with health indicators, also in school-aged children and adolescents, which found that other health factors than adiposity have been under-researched to date [72].

The current review also found that most research has focused on either sedentary bouts/breaks or MVPA bouts, with only two of the 29 studies examining LPA patterns [55, 56]. Whilst some studies acknowledge the importance of other intensities by adjusting for them, none have examined multiple intensity patterns together (e.g., using cluster analyses, compositional analyses [73, 74]). Young people do not engage in activity intensities in isolation and there are many ways to accumulate time through different combinations of sedentary behaviours and physical activity [73, 74]. Specifically, when a child is engaging in one activity intensity, they cannot be engaging in another activity intensity. It is these combinations of activity patterns (e.g., how often and with what intensity should you break up sitting) that have been understudied [73, 74]. While there is already some recognition of conceptualising sedentary time and physical activity (including all intensities such as LPA and MPA) together as part of a continuum in total volume research [72, 74], most studies have only investigated independent effects of patterns for a single intensity. Consequently, it is not known whether differential associations are observed between combined activity patterns and cardio-metabolic risk factors.

Future directions

Whilst this review highlights that there is emerging interest in examining associations between activity patterns and cardio-metabolic risk factors in children and adolescents (100% of studies were published from 2009 onwards, and 45% from 2015 onwards), few studies have consistently examined the same activity patterns with cardio-metabolic risk factors. This made it impossible to recommend how children and adolescents should accumulate their activity to benefit their cardio-metabolic health. To aid comparability between studies, it would be beneficial for future research to consider standardised pattern definitions and assessments of activity patterns (e.g., accepted methodology and terminology for assessing sporadic and prolonged activity patterns). This would enable studies to replicate these assessments in different populations, thus building the evidence base in relation to whether activity patterns impact cardio-metabolic health outcomes in children and adolescents. More evidence is also needed for cardio-metabolic risk factors other than adiposity, such as blood lipids. This research should include the full activity spectrum from sedentary to vigorous, and the composition of different activity patterns (e.g., combined activity patterns of bouts and breaks of multiple intensities). As it is possible that there will be differing association with cardio-metabolic health depending on the combination of activity patterns across the activity spectrum [73, 74], it is important that future research also examines associations of these combinations and cardio-metabolic health in children and adolescents. This will contribute to the understanding of activity patterns in children and adolescents, provide knowledge as to whether existing physical activity and sedentary behaviour guidelines need to be refined to include recommendations relating to specific activity patterns, and enhance the development of targeted interventions to benefit health outcomes.