Accelerometer-determined peak cadence and weight status in children from São Caetano do Sul , Brazil

The purpose of this study was to determine the relationship between peak cadence indicators and body mass index (BMI) and body fat percentage (BF%)-defined weight status in children. The sample comprised 485 Brazilian children. Minute-by-minute step data from accelerometry were rank ordered for each day to identify the peak 1-minute, 30-minute and 60-minute cadence values. Data were described by BMI-defined and bioelectrical impedance-determined BF% weight status. BMI-defined normal weight children had higher peak 1-minute (115.5 versus 110.6 and 106.6 steps/min), 30-minute (81.0 versus 77.5 and 74.0 steps/min) and 60-minute cadence (67.1 versus 63.4 and 60.7 steps/min) than overweight and obese children (p<.0001), respectively. Defined using %BF, normal weight children had higher peak 1-minute (114.5 versus 106.1 steps/min), 30-minute (80.4 versus 73.1 steps/min) and 60-minute cadence (66.5 versus 59.9 steps/min) than obese children (p<.0001). Similar relationships were observed in boys; however, only peak 1- minute cadence differed significantly across BMI and %BF-defined weight status categories in girls. Peak cadence indicators were negatively associated with BMI and BF% in these schoolchildren and significantly higher among normal weight compared to overweight or obese children.


Introduction
Childhood obesity is an important public health problem that has recently increased dramatically in prevalence in both developed and developing countries 1 .A recent systematic review concluded that higher levels of regular physical activity assessed objectively by accelerometers likely protects against obesity in children and adolescents 2 .The use of accelerometers to measure physical activity generally results in stronger associations with overweight and obesity than the use of questionnaires in children 3 .For example, Ferrari et al. 4 reported a negative association between objectively measured moderate-to-vigorous physical activity (MVPA) with body composition variables (body mass index (BMI) and body fat percentage (BF%)) in Brazilian children.
Researchers have used step counts to describe children's and adolescents' daily ambulatory physical activity levels 5,6 .However, steps/day has been criticized because it only provides a total volume of physical activity without information about the intensity.This limitation can be overcome with the use of accelerometers that have the capability of storing minute-by-minute step accumulations, which can be used to quantify stepping cadence (steps/min) as it is naturally expressed in free-living, indicative of a full range of step accumulation patterns and stepping rates 7,8 .
Originally investigated by Hoshikawa et al. 9 , cadence is associated with ambulatory speed 10 and is a proxy indicator of intensity of ambulatory activity [11][12][13] .A recent review has introduced the concept of using cadence as a way to study free-living ambulatory behavior 14 .Some studies have used accelerometer step counts to not only describe the total amount of daily ambulation but also different aspects of children's cadence [15][16][17] .In addition to analyzing cadence to describe movement it is also possible to measure and analyze patterns of peak cadence for different time increments.Peak cadence indicators for children include peak 1-minute (defined as the steps/min recorded for highest single minute in a day), peak 30-minute (average value for top 30 ranked minutes) and peak 60-minute (average value for top 60 ranked minutes) cadence values and can be considered indices of "best natural daily effort" 18,19 .
The relationship between peak cadence and health related variables has been examined in children from developed countries [16][17][18] .Using data from the 2005-2006 U.S. National Health and Nutrition Examination Survey (NHANES), Barreira et al. 18 reported negative associations between cardiovascular disease (CVD) risk factors (in particular, HDL-C, blood pressure, and BMI) and each of the peak cadence indicators in children.Peak 60-minute cadence was lower across groups of children with higher numbers of CVD risk factors (76 steps/min in those without any risk factors versus 67 steps/min in those with ≥ 2 risk factors) 18 .In addition, Gardner et al. 16 reported that children, adolescents, and young adults with metabolic syndrome walked at lower daily average cadence than those without metabolic syndrome (27 versus 30 steps/min), and they had lower cadences for continuous durations of 60 min (31 versus 38 steps/min), 30 min (42 versus 50 steps/min), and 1 min (102 versus 110 steps/min).Although popular in use, there are relatively few studies that have used accelerometry to study children in developing countries like Brazil 4,20,21 .The relationship between peak cadence indicators and obesity has not been studied in Brazilian children.Therefore, the purpose of this study was to determine the relationship between peak cadence indicators and weight status (defined by BMI and by BF%) in children from São Caetano do Sul, Brazil.

Methods
Data collection was conducted as part of the International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE).The primary aim of ISCOLE was to investigate the influence of behavioral settings and the physical, social, and policy environments on observed relationships between lifestyle characteristics and weight status in approximately 500 children from each of the included 12 countries (Australia, Brazil, Canada, China, Colombia, Finland, India, Kenya, Portugal, South Africa, United Kingdom, and United States) representing all major world regions.Details of the ISCOLE protocol are provided elsewhere 22 .
The analysis herein focuses only on the data collected in the city of São Caetano do Sul, located in the state of São Paulo, Brazil, with a land area of 15.3 km 2 and a subtropical climate.The population of the municipality in 2013 consisted of 149,263 inhabitants, including 1,557 children (812 boys and 745 girls) 10 years of age 23 .The city is characterized as a service economy 23 and has the best Human Development Index (HDI) in Brazil (0.86) according to the United Nations Program for Development 24 .
A random cluster sample of 564 5 th grade Brazilian children (277 boys and 287 girls), aged 9-11 years old, was assessed.After accounting for exclusionary criteria (non-valid accelerometer data as defined below, missing BMI or BF%, the final sample comprised 485 children (238 boys and 247 girls)).
Data were collected during the school year from March 2012 to April 2013.All assessments were done during a full week per school.All data collection and management activities were performed and monitored under rigorous quality control procedures, implemented by the ISCOLE Coordinating Center, as described in detail previously 22 .Prior to participating, children and at least one of their parents/legal guardians were asked to sign the Instrument of Consent according to Resolution 196/96 of Brazil's National Health Council.Ethical approval was obtained from the Pennington Biomedical Research Center Institutional Review Board and Federal University of São Paulo, Brazil.
Random lists of public and private elementary schools in the region were generated, and schools were selected from each list at a ratio of 4 (public) to 1 (private).This 80% public to 20% private schools ratio was purposely implemented to maximize socioeconomic status distribution.If a school refused to participate in the study, it was replaced by the next school on the list, maintaining the same public to private school ratio.A random sample of 25-30 children was selected to participate per school with a stipulation that each sex comprised 50% of the selected sample.
The ActiGraph GT3X+ accelerometer (Acti-Graph LLC, Pensacola, United States) was used to assess step-defined physical activity.The accelerometer was worn at the waist on an elasticized belt, on the right mid-axillary line.The participants were encouraged to wear the accelerometer 24 hours/day for at least 7 days (plus an initial familiarization day and the morning of the final day), including 2 weekend days.To be included in this analysis, children had to have valid accelerometer data (defined as ≥4 days, including at least one weekend day, with ≥10 hours of wake wear time per day) 25,26 .Overall, 84.7% of the eligible sample met these criteria.The number of valid days ranged from 4-7, averaging 6.7 days per participant.
Following the final day of data collection, staff went into the school and retrieved the accelerometers.The research team verified the data for completeness using the most recent version of the ActiLife software (version 5.6 or higher; ActiGraph, Pensacola, United States) available at the time.Nine participants who did not provide adequate data during initial monitoring wore the accelerometer for a second week to ensure that the minimal data requirements were met.The step data were processed with the default filter so as not to over-estimate step counts using the low frequency extension 27,28 .
Due to the 24-hour wear protocol, the nocturnal total sleep episode time was identified and excluded from any analysis of steps/day and physical activity using a published algorithm 29 .Daily minutes of wake wear time were then ranked in descending order of steps/min 17 .The value for the highest ranked 1 minute of steps/ min represented the peak 1-minute cadence.Peak 30-minute (highest 30 minutes) and peak 60-minute (highest 60 minutes) cadence represented an average steps/min accumulated during the highest, but not necessarily consecutive, min/ day.Each of these peak cadence indicators was averaged across all valid days 6,30 .
Height, weight and BF% measures were obtained according to standardized ISCOLE procedures 22 .Height was measured using a Seca 213 portable stadiometer (Hamburg, Germany) 31 .Weight and BF% were measured using a portable Tanita SC-240 body composition analyzer (TANITA Corporation, Japan) 32 .Each child was measured twice and, when necessary, a third measurement was taken if the difference between the previous two was outside the permissible range for each measure and its replica (0.5 cm for height, 0.5 kg for weight, and 2% for BF%).The two closest measurements were averaged, and the mean value of each measured variable was used for analysis.
BMI was calculated using the standard formula [weight(kg)/height(m) 2 ] and thereafter, BMI z-scores (calculated based on sex-and age-specific growth reference data from the World Health Organization (WHO) for children and youth) were further categorized as normal: ≤ 1 SD; overweight: >+1 SD to 2; and obese: >+2 SD 33 .
We also categorized BF% according to sexand age-specific cut-points.Specifically, children were categorized as normal weight (< 85 th percentile of sex-specific reference data from children in the United Kingdom) 34 , overweight (≥ 85th to <95 th percentiles) and obese (≥ 95 th percentile).
Descriptive statistics, including means, confidence interval (95% CI) or frequencies as appropriate were calculated.Multi-level linear regression models were used to examine the associations between peak cadence indicators and BMI and BF%.Peak cadence indicators across different levels of weight status defined by BMI and BF% (normal weight, overweight and obese) were analyzed in a model that included sex and the sex-by-weight status interaction.Least-squared means were compared across levels of BMI and BF% for the entire sample as well as separately by sex.Multi-level linear regression models were also used to examine the associations between peak cadence indicators and continuous measures of BMI and BF%; models were adjusted for sex and accounted for school (to allow for clustering at the school level).Statistical Analysis System (SAS, version 9.3) was used for data analyses and p < 0.05 was adopted as the significance level 35 .

Results
The frequency of excess BMI and BF% status is presented in Table 1.Based on BMI, most children were classified as normal weight (54.6%) followed by overweight (23.1%) and obese (22.3%).When analyzed separately by sex, 45.8% of boys and 45% of girls were overweight or obese according to their BMI-defined weight status.Most children (67%) were classified as normal weight by BF%, followed by overweight (12.8%) and obese (20.2%).By sex, 33.6% of boys and 32.4% of girls were classified as overweight or obese by BF%.
Compared to girls, boys consistently achieved significantly higher values for each of the peak cadence indicators (Figure 1).
Table 2 presents peak cadence indicators across categories of BMI.For the total sample, there were significant differences in each of the peak cadence indicators between BMI-defined weight status categories.Normal weight children achieved significantly higher mean values for each of the peak cadence indicators compared with overweight and obese children (Table 2).
The same relationships were observed when the analysis was focused only on boys: there were significant differences in all peak cadence indicators across BMI-defined weight status categories and normal weight boys had higher mean values obese boys.In contrast, only the peak 1-minute cadence differed significantly across BMI-defined weight status categories in girls.
In the normal weight category, boys achieved significantly higher mean values than girls in all peak cadence indicators, and overweight boys had significantly higher mean values than overweight girls in the peak 60-minute cadence indicator category (Table 2).There were no significant sex differences in the obese category.
Table 3 presents the peak cadence indicators across categories of BF%-defined weight status.All peak cadence indicators had mean values that differed significantly across BF% weight status categories.Normal weight children achieved significantly higher mean values than obese children in all peak cadence indicators and overweight children had significantly higher values than obese children for the peak-1 minute indicator (Table 3).
In boys, there were significant differences among the BF% weight status categories for all peak cadence indicators; normal weight boys had significantly higher mean values than obese boys in all peak cadence indicator categories and significantly more than overweight boys in the peak-60 minute cadence indicator category.Normal weight girls achieved significantly higher mean values than obese girls for the peak 1-minute indicator only (Table 3).In the normal weight category, boys accumulated significantly higher mean values than girls in all peak cadence indicators; no sex differences were observed in the obese or overweight categories (Table 3).
Table 4 presents the results of the multi-level regression analyses describing the association between each of the peak cadence indicators and BMI and BF% separately.There were negative significant associations between BMI and all peak cadence indicators as well as between BF% and all peak cadence indicators when adjusted for sex and school, indicating that this association is independent of sex and school.

Discussion
The aim of this study was to determine the relationship between peak cadence indicators and weight status in children from São Caetano do Sul, Brazil, averaging 10 years of age.We demon-

Peak cadence indicators
Steps/min strated significant differences between each of the peak cadence indicators and BMI or BF%-defined weight status in the total sample and in boys when analyzed separately.Only the peak 1-minute cadence differed significantly across BMI and BF% weight status categories in girls.Boys consistently achieved significantly higher mean values for each of the peak cadence indicators compared to girls.We found significant negative associations between each of the peak cadence indicators and BMI or BF% continuous variables when controlling for sex and school.This study supports previous research that has shown negative relationships between peak cadence indicators and weight status 17,36 .For example, Jago et al. 36 divided American boys aged 11-15 years into groups of normal weight (BMI < 85 th percentile) and at risk of being overweight (BMI ≥ 85 th percentile) and reported that normal weight children achieved significantly higher mean steps/min values during continuous slow walk (121.0 versus 110.1 steps/min), fast walk (129.4 versus 122.5 steps/min), and running (168.6 versus 153.2 steps/min) compared with children at risk of being overweight.Our results demonstrated that normal weight boys (defined by BMI as ≤ 1 SD) had higher peak 1-minute (118.8 versus 107.1 steps/min), 30-minute (86.0 versus 74.9 steps/min) and 60-minute (72.4 versus 62.0 steps/min) free-living cadence than overweight children (p < .0001).
Barreira et al. 17 analyzed the 2005-2006 NHANES data and showed that normal weight (defined as < 85 th percentile BMI) boys and girls (combined) had higher peak cadence (peak 1-minute, 30-minute and 60-minute) than those defined as obese (≥ 95 th percentile BMI).For example, for peak 60-minute cadence, children with normal weight BMI had a mean of 73 steps/ min and obese children had a mean of 67 steps/ min; however, in the peak 30-minute and peak 60-minute cadences there were no mean differences between overweight and obese.Our study reported that children classified as overweight/ obese had lower values for all peak cadence indicators than normal weight children by BMI status and we found similar differences in peak 60-minute cadence between BMI-defined nor-  mal weight (67.1 steps/min) and obese children (60.7 steps/min).We found significant negative associations between weight status and each of the peak cadence indicators.Our decision to link peak cadence indicators to sex-specific weight status cut points (BMI and BF%) 33,34 discriminating normal weight and overweight/obesity is justified based on the overwhelming evidence of an increasing world-wide obesity epidemic 1 and further supported by the statistical differences in peak cadence indicators identified herein between sex and BMI and BF%-defined weight status categories.There were significant differences in peak cadence indicators across weight status (BMI: normal weight versus overweight and obesity; BF%: normal weight versus obesity) categories for boys, but in girls only the peak 1-minute cadence indicator differed between normal weight and obesity (BMI and BF%).Girls appeared to be homogeneously lower in their "best natural effort" regardless of weight status category, especially for the peak 30-minute and 60-minute cadence indicators, which are shaped not only by intensity of effort but also by relative "persistence" of this behavior, whereas peak 1-minute is not similarly shaped by such persistence.
In the present study, boys had significantly higher mean values for each of the peak cadence indicators than girls.Our findings agree with recent studies that boys accumulate more steps/ day than girls 17,37,38 .Barreira et al. 17 demonstrated differences between the sexes only for peak 60-minute cadence (73 steps/min for boys versus 70 steps/min for girls).The authors found no significant differences in peak 1-minute or peak 30-minute cadence indicators by sex.Recently, Barreira et al. 38 reported sex-and-age specific normative values of peak 60-minute for children and adolescents from United States using data from the 2005-2006 NHANES and found that boys take 74.5 steps/min and girls take 72 steps/ min.In a review article attempting to answer how many steps/day are enough for children and adolescents, Tudor-Locke et al. 37 indicated that, on average, boys and girls take 12 000-16 000 and 10 000-13 000 total steps/day, respectively.
Findings presented here suggest a strong relationship between peak cadence and weight status independent of sex and school.We reported negative associations between each of the peak cadence indicators studied and both BMI and BF%.Similarly Gardner et al. 16 reported sex-adjusted differences in daily average cadence related to presence or absence of metabolic syndrome (14.9 versus 13.6 steps/min) in children, adolescents, and young adults.The lower values for peak cadence indicators in obese children should also be related to lower daily energy expenditure of physical activity 39 , setting the stage for long-term positive energy balance and exacerbating further increases in body weight and fatness.Collectively, our results support the notion that intensity of physical activity expressed as peak cadence indicators is an important factor in distinguishing between normal weight, overweight and obese children.This is further supported by studies showing that time spent in MVPA is protective against developing weight status in children 2,40,41 .
Weight status is an important correlate of daily ambulation patterns in children.BMI and BF% were both significantly and negatively associated with each peak cadence indicator.These results indicate that participants with high levels of body fat ambulate at lower peak cadences than those with less body fat 16 .Our findings agree with others that have reported a strong association between obesity and accelerometer-determined physical activity 2,40 .
The strengths of this study include the objective measurement of peak cadence indicators and the measurement of body fat using bioelectrical impedance in Brazilian children.These techniques and approaches are rare in Brazil where most previous research has relied upon indirect measures of physical activity collected by questionnaires [42][43][44] .This is the first study that has used multi-level modeling to examine the relationship between peak cadence indicators and weight status in Brazilian children; however, there are some limitations of this study that must be acknowledged.The cross-sectional nature of this study prevents determination of whether these associations are in some way causal or whether physical activity may be a marker of some other lifestyle factors (e.g., dietary factors or socioeconomic status) that may influence these weight status variables.Another limitation is that step-counting devices, including waist-worn accelerometers, do not quantify nonambulatory physical activity (e.g., resistance training, swimming) that could also be considered moderate intensity, and therefore may underestimate the total amount of daily physical activity performed to some extent.This is a descriptive study of step accumulation patterns of children from São Caetano do Sul, Brazil.Since this analysis shows a relationship between these peak cadence indicators and BMI and BF%, it is important now to move beyond the descriptive epidemiology to examine their relationship with other health parameters using cross-sectional, longitudinal, and intervention study designs.

Conclusion
This analysis of the ISCOLE accelerometry data represents the first examination of the relationship between peak cadence indicators and BMI and BF%-defined weight status in children from São Caetano do Sul, Brazil.
The present study provided evidence of significant negative associations between BMI as well as BF% and peak cadence indicators (peak 1-minute, 30-minute and 60-minute) among 10-year old Brazilian children, independent of sex and school.We reported significant differences in each of the peak cadence indicators across BMI or BF%-defined weight categories in the total sample and in boys.Only the peak 1-minute cadence differed significantly across BMI and BF% weight status categories in girls.Boys had higher values for all peak cadence indicators than girls in the normal weight category.
Further research is needed to continue to develop means of gathering more comprehensive data in order to better elucidate the full nature of the correlates of obesity.These reference data are novel and important and can be used for surveillance, tracking, comparison, screening, intervention, and evaluation purposes.

Collaborations
GLM Ferrari conceived, designed, and implemented the study, collected and helped to write and revise the manuscript; T Araújo, LC Oliveira and VKR Matsudo helped to implement the study and to write the manuscript; E Mire performed statistical analyses and interpreted the data; T Barreira was responsible for the data collection, helped implement the study, and helped to write the manuscript; C Tudor-Locke interpreted the data, and helped to write and revise the manuscript; PT Katzmarzyk was responsible for coordinating the study and contributed to the intellectual content.All authors contributed to the study design, critically reviewed the manuscript and approved the final version.

Figure 1 .
Figure 1.Peak cadence indicators by sex in Brazilian children.*Significant difference between boys and girls.

Table 1 .
Frequency (%) of BMI and BF%-defined weight status by sex in Brazilian children.

Table 2 .
Descriptive (mean and 95% CI) analysis of peak cadence indicators according to BMI-defined weight status and sex in Brazilian children.

Table 3 .
Descriptive (mean and 95% CI) analysis of peak cadence indicators of according BF% status and sex in Brazilian children.: body fat percentage.a Trend in least-squared means across BMI-defined weight status categories; b significant difference between boys and girls (p < .05);c significant difference between normal weight and overweight (p < .05);d significant difference between normal weight and obese (p < .05);e significant difference between overweight and obese (p < .05). BF%

Table 4 .
Adjusted analyses between peak cadence indicators, BMI and BF% in Brazilian children.