The objective of this study was to examine the association between accelerometer-measured sedentary time and health risk in children. Our analysis supports previous studies that found few or no significant associations between accelerometer-measured sedentary time and health risk in children [1, 11–16]. This study is novel because it included a wider range of sedentary time variables than what has been previously considered that characterize the timing and patterning of how the sedentary time is accumulated. Further, the sedentary pattern variables were designed to be more reflective of real-world sedentary behaviour. For example, a limited number of short transitions into light activity were allowed to reflect real-life situations where individuals are sedentary for long periods but move around occasionally (i.e., to go to the washroom or answer the phone). Despite the inclusion of more comprehensive sedentary pattern variables, this study found few significant relationships with health risk and the associations observed were limited to boys aged 11–14 years.
In theory, excessive sedentary time is associated with negative health outcomes [4, 37] and self-reported screen time is associated with elevated health risk in children [1–3]; however, the way we currently measure sedentary time with accelerometers does not consistently support this link. To date, the research linking accelerometer-measured sedentary time with health risk among children and youth has been mixed. It is therefore unclear whether a relationship exists only in some populations or if differences in analytical approaches explain the inconsistencies observed. There appears to be more evidence supporting a lack of relationship between accelerometer-measured sedentary time and health risk in children and youth [1, 11–16] than there is supporting a relationship [8–10]. Adjustment for MVPA appears to attenuate significant associations between accelerometer-measured sedentary time and health risk [13, 16], suggesting that MVPA is more powerful than total sedentary time at explaining the variance in health risk in children and youth. In our unadjusted regression models, sedentary time was associated with BMI and waist circumference in boys aged 6 to 14 and girls aged 6 to 10 years; however, after adjustment for MVPA, these associations remained significant only in 11–14 year old boys.
In 2008, Healy and colleagues published a paper that reported a significant association between number of daily breaks in accelerometer-measured sedentary time and lower metabolic risk in adults . This work led researchers to question whether it is the pattern of how sedentary time is accumulated, rather than simply the total volume of sedentary time, which matters for health. Do frequent interruptions in sedentary time attenuate the health risk that sedentary time imposes? Does this relationship apply in both children and adults? Carson and Janssen found no significant associations between breaks in sedentary time or prolonged bouts of sedentary time lasting 30 minutes with cardio-metabolic risk factors in a large sample of American children . Number of breaks in sedentary time was only associated with waist circumference in 11–14 year old boys in the present analysis. We included an additional layer of complexity by examining sedentary time, breaks and prolonged bouts of sedentary time during periods of discretionary free time: weekends and after school. We hypothesized that sedentary time accumulated during these periods would better discriminate between children engaging in healthy and unhealthy levels of sedentary behaviour when compared to simply examining overall sedentary time. We observed no significant associations between the patterns of sedentary time accumulated on weekends and health risk in children; however, some relationships emerged when we examined sedentary time accumulated during the after school period. Interestingly, we only observed significant associations in boys aged 11 to 14 years of age when the regression models were adjusted for age, MVPA and accelerometer wear time.
It is difficult to speculate why we observed significant findings in boys and not girls. It is possible that more overweight and obese boys in this sample were engaging in prolonged bouts of sedentary time after school, a finding consistent with previous research that has found that boys spend considerably more time in specific sedentary behaviours such as video game playing [38–40]. Average daily sedentary time and weekend sedentary time did not differ between boys and girls while sedentary time accumulated after 3 pm on weekdays was higher in boys compared to girls (277 vs. 266 minutes). Significant differences between boys who are not overweight/obese and overweight/obese boys were observed in the sedentary variables; however, no such differences were observed in girls. For example, there was virtually no difference in average daily sedentary time between girls who are not overweight/obese versus those who are (524 vs. 525 min∙d-1) while a more marked difference existed between boys who are not overweight/obese versus those who are (500 vs. 528 min∙d-1). In Figure 1, a distinction between boys who are not overweight/obese and those who are can be observed across all bout lengths; however the difference is only statistically significant when the bout length is at least 80 minutes long. By comparison, no difference is noticeable by overweight/obesity status in girls and there is more cross-over in the error bars in girls (Figure 2). Similarly, no significant associations emerged in girls in the regression analyses while in 11–14 year old boys, prolonged bouts of sedentary time lasting at least 80 minutes, accumulated during the after school period were associated with both BMI and waist circumference. Explaining why significant associations were observed in 11–14 year olds boys but not those who were 6–10 or 15–19 years is not easy. In a large sample of US children, Sisson and colleagues observed an increase in screen time with age from 2 to 15 years . In the Health Behaviour and School Aged Children Survey, the Canadian data show that screen time increases from age 11 to 15 years  with the peak occurring in grade 9 (approximately 14 years) . These large data sets suggest that the 11–14 year old age group may be an age range where screen time habits change significantly, thus increasing the amount of variation (and likelihood to find significant associations) in this variable.
The lack of evidence linking accelerometer-measured sedentary time with health risk in children is counter-intuitive given the consistent observation that screen time, a key contributor to total sedentary time, is associated with health risk [1–4]. One of the fundamental differences between self-reported screen time and objectively measured sedentary time is that the former is capturing one specific activity while the latter is capturing screen time in addition to many other sedentary behaviours. Much of the time accumulated as “sedentary” represents normal aspects of day-to-day life therefore capturing every minute in a day that is sedentary, as accelerometers do, may dilute the associations between specific sedentary behaviours (e.g., watching television) and health risk. It is possible that some forms of sedentary behavior (e.g., screen time, long car or bus travel) are associated with negative health outcomes while other forms of sedentary behavior (e.g., eating, reading, resting, socializing etc.) are not. Similarly, data reduction procedures used in accelerometry analysis (e.g., 10 hour wear time criteria) were developed to accurately capture MVPA and whether they are appropriate for sedentary behaviour research questions is unknown. For example, it has been suggested that wear time has a disproportionate impact upon estimates of sedentary time compared with MVPA . Teasing out which sedentary behaviours beyond screen time are associated with negative health outcomes represents an important area for future research.
The sedentary behaviours that are of known public health concern in children and youth (e.g., excessive levels of screen time) typically last for extended periods of time (i.e., up to several hours at once). This reality was the motivation behind the way prolonged bouts of sedentary time were defined in the present analysis. Had we used a strict definition of what ended about (i.e., any transition out of sedentary) then our longest bout length would have been very short (e.g., 10 minutes) and thus not representative of one of the key sedentary behaviours that we were interested in capturing. The allowance of a modest amount of light intensity movement within the prolonged sedentary bouts was therefore purposeful and allowed much longer bout lengths to be examined (up to 2 hours). Number of breaks per day, also assessed in this analysis, is an important aspect of sedentary behaviour patterns. Given that we and others  have not consistently observed significant associations between number of breaks per day and health risk, it is important to look at alternative pattern variables such as prolonged bouts. Further, the extension of bout length in the present analysis was important to build off the only other published work that examined prolonged bouts up to 30 minutes in children and youth .
It is possible that the true health effect of sedentary time is attenuated by limitations with the data and analysis. Possible limitations that dilute the possibility of observing a true relationship include: i) the cross-sectional nature of the data, ii) non-response bias, iii) the possibility that the findings in 11–14 year olds boys reflect Type 1 error. As described in the methodology, the non-response bias is adjusted for in the data. We attempted to minimize the likelihood of Type 1 errors in the regression analyses by adjusting the p-value for significance from .05 to .006. Other limitations include the lack of ability to confirm precisely when children were finished school. We examined the period after 3 pm on weekdays  based on the assumption that most kids would finish school sometime between 2-4 pm. Accelerometers are limited in their ability to capture postural changes (i.e., cannot differentiate between sitting and standing) and are therefore limited in their ability to measure sedentary before as well as other tools which encompass an inclinometer in addition to an accelerometer. No significant associations were observed between sedentary time variables and blood pressure or non-HDL cholesterol and this may be due to it likely being more difficult to detect meaningful differences in biomarkers in children and youth than adults because younger people are more distal to pathophysiological developments. A similar examination on a population of high-risk children (e.g., overweight or obese or with a family history of cardio-metabolic disease) may lead to different findings as these children would be more likely to exhibit abnormalities in blood markers and blood pressure. Finally, examination of interaction and confounding effects was limited because the number of variables (including interaction terms) that can be tested within the CHMS data set is limited by the available degrees of freedom.