Scand J Med Sci Sports 2015: 25: 706–715 doi: 10.1111/sms.12293
© 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Effects of extra school-based physical education on overall physical fitness development – the CHAMPS study DK C. T. Rexen1, A. K. Ersbøll2, N. C. Møller1, H. Klakk1,3 N. Wedderkopp1,4,5, L. B. Andersen1,6 1
Centre of Research in Childhood Health, Institute of Sports Science and Clinical Biomechanics, University of Southern Denmark, Odense M, Denmark, 2National Institute of Public Health, University of Southern Denmark, København K, Denmark, 3University College Lillebaelt, Odense M, Denmark, 4Institute of Regional Health Services Research, University of Southern Denmark, Odense M, Denmark, 5Sport Medicine Clinic, Orthopedic Department, Hospital of Lillebaelt, Middelfart, Denmark, 6Department of Sports Medicine, The Norwegian School of Sport Sciences, Oslo, Norway Corresponding author: Christina Trifonov Rexen, Centre of Research in Childhood Health, Institute of Sports Science and Clinical Biomechanics, University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark. Tel: +45 2671 9691, E-mail: [email protected] Accepted for publication 19 June 2014
First, this study aimed to investigate if four extra physical education (PE) lessons per week improved children’s development in physical fitness. Second, to investigate if the extra PE lessons improved development in physical fitness for children with lower levels of fitness at baseline. This study was a longitudinal controlled school-based study. The study population consisted of 10 Danish public schools with children in preschool to fourth grade (cohorts 0–4) with 2.5-year follow-up. Six schools had extra PE and four schools had normal PE. In total 1247 children were included (normal PE = 536, extra PE = 711). Development in fitness was analyzed using a composite z-score from six fitness tests. Multilevel mixed-
effects linear regression was used to examine the association between school type and development in fitness. Extra PE increased the total development of composite z-score units among children enrolled in cohort 4 and borderline in cohort 3 with 1.06 (95% confidence interval 0.48–1.65) and 0.52 z-score units (−0.06 to 1.09), respectively. Children in the lower 50 percentiles increased their development with 0.47 (0.08–0.85) z-score units. Extra PE in schools improved development in fitness for cohort 4 and borderline for cohort 3 among all children. Extra PE improved fitness development across all cohorts among children with low fitness levels.
A polarization in fitness and other health-related measures among school-aged youth has been identified over the past decades (Wedderkopp et al., 2004; Dyrstad et al., 2012). It is not yet clear whether this can be restored through school-based interventions. Furthermore, the optimal quality, quantity, and intensity in school settings have yet to be discovered, although multicomponent interventions have already been shown to have a positive effect on the overall level of physical activity (PA) and possibly also a positive effect on fitness among school-aged youth (Kriemler et al., 2011). School-based interventions evaluating physical fitness components have identified positive effects on cardiorespiratory fitness (Hansen et al., 1991; Resaland et al., 2011) and several physical fitness components (Sollerhed & Ejlertsson, 2008). However, school interventions with no significant effects on physical fitness have also been reported (Bugge et al., 2012; Magnusson et al., 2012; Christiansen et al., 2013). A review by Kriemler et al. (2011) found positive effects of schoolbased interventions on fitness in 6 out of 11 studies,
suggesting that school setting could be an important arena for restoration of the current negative trend in fitness among healthy youth. The question of how much extra physical education (PE) is needed to document longitudinal effects on fitness is not easily answered. A review by Strong et al. (2005) suggests at least 60 min of moderate-to-vigorous daily PA for school-aged youth to accomplish effects on aerobic fitness, muscular strength and endurance, and other health-related effects of regular PA. Motor performance or fundamental movement skills are considered the foundation of more complex and specialized movement sequences required for participation in organized sport as well as many non-organized activities during free play in childhood and adolescence (Clark & Metcalfe, 2002; Gallahue & Ozmun, 2006). Most movements involved in an active lifestyle have elements of both gross and fine motor components (Payne & Isaacs, 2002). Fundamental movement skills are developed in childhood and subsequently refined into more complex context- and sport-specific skills (Clark, 2005;
706
Effects of extra physical education in school Stodden et al., 2008) during early and middle school years (Gallahue & Ozmun, 2006). Therefore, it seems relevant to combine testing of motor performance (quantitatively assessable movement skill; Gallahue & Ozmun, 2006) with measures of cardiorespiratory fitness in children when evaluating the effects of extra PE lessons. Furthermore, school-based interventions aiming to increase PA and fitness are thought to be effective and universally applicable because PE in schools is compulsory in most countries and includes the least active children (Kriemler et al., 2011) at a developmentally important age. Children with low levels of fitness are particularly interesting from a health-related perspective (Brage et al., 2004; Andersen et al., 2006). Objectives First, the objective of this study was to investigate if six weekly PE lessons were sufficient to improve physical fitness development in schoolchildren when compared with the traditional two weekly lessons in normal schools. Second, to investigate if children with low levels of fitness at baseline improved their development of physical fitness in schools with extra PE compared with normal schools. Both aims were answered using intention-to-treat analyses. Methods Study population The CHAMPS study DK is a longitudinal study. The initial concept was brought forth at municipality level by local authorities and extended by researchers to include control schools. The study design can therefore be described as a quasi-experimental study evaluating a natural experiment (Craig et al., 2012). In Denmark, children attend primary public schools from the age of 6 (preschool or grade 0) to the age of 16 or 17 (9th or 10th grade). For smaller schools, this may involve school change at grade level 7. All public schools in Denmark are under the influence of governmental reform at a national level. This reform is adapted to match resources and priorities within each municipality by local authorities. Finally, detailed planning is made at school level by the school itself. In the CHAMPS study DK, all 19 primary schools in the municipality of Svendborg (Southern Funen, Denmark) were invited to participate in the project as schools with extra PE (270 min/week divided over at least three sessions per week). Ten schools agreed to participate, but only six schools were willing to prioritize the financing of the extra PE sessions. To study the effect of extra PE in schools, the municipality was asked to obtain matched control schools according to school size, urban/rural area, and socioeconomic position among the 13 schools that continued the normal curriculum of 90-min PE per week (one session). This resulted in a total of n = 773 and n = 734 invited children in schools with extra PE and normal PE, respectively. Parents and children were kept unaware of the initiation of this project until after the deadline for school choice, thereby avoiding influence by parents on the choice of school. All 10 schools in the study accommodated all grade levels within the study period (0–7th grade). Detailed description of the study protocol is published elsewhere (Wedderkopp et al., 2012). The children were followed for 3 years. At baseline (August 2008) 1222 children
Fig. 1. Flow chart of participant at various time points. The number and school type [normal or extra physical education (PE) school] of drop-in or out are shown. 1 Eighty-one percent of the total invited participants agreed to participate (at test occasion 1). 2In total, 71% of the invited participants at schools with normal PE agreed to participate. Ninety-one percent of the children at extra PE schools agreed to participate. From the total of 99 dropouts, 83 were due to change of school. participated (by school type: normal PE: n = 522, extra PE: n = 700). The project was continuously open for new participants, and a total of 99 dropout and 108 drop-in were observed during the 2.5-year follow-up (Fig. 1).
Socioeconomic status There were no significant differences in household income between the two school types. However, household income was 15% lower in non-participating schools. There was no significant difference of parental education between school types or between participating vs non-participating schools in the municipality.
Ethics The study was carried out in accordance with the Declaration of Helsinki and was approved by the local scientific ethics committee (ID S20080047) and registered in the Danish Data Protection Agency (J.nr. 2008-41-2240). Parents of the children gave written informed consent and the children gave verbal consent.
PE lessons in schools with extra PE To ensure optimal quality of the PE lessons, PE teachers and pedagogues participated in a 40-lesson course in “Age-related concepts of training” organized by Team Denmark in collaboration with the faculty of education at the University College Lillebaelt December 2008–April 2009. The goal was to
707
Rexen et al. ensure targeted and responsible training for the growing individual (Pryce et al., 2005). Furthermore, a handbook was provided by Team Denmark with inspiration for the content of PE lessons, practical suggestions, and exercises. The handbook is published by Bach and Eiberg (2010). Overall, the concept involves focus on gradual development of all-round fundamental motor skills for the prepubertal children through organized playing activities involving training of movement skills, e.g., running, hopping and jumping, throwing and catching, kick and dribble, orientation skills and sensation of positions, simple balance, rhythm, coordination, reaction, and timing. As the children enter puberty, the activities and training becomes more complex and advanced and approaches more sports-specific skills, e.g., activities with repeated hops and jumps, change of direction, speed, and more advanced ball games including ball-oriented timing and agility.
Physical fitness testing Participating children were tested in total five times during a period of three school years. Four tests were performed with half a year interval during the first 2 years and again 1 year later at the end of the 3-year period. Baseline data were collected in September–October 2008. The quantitative fitness testing consisted of five motor performance tests and one aerobic performance test. These tests were: • Short shuttle run test from the Eurofit test battery (Adam et al., 1991) • Handgrip strength test from the Eurofit test battery (Adam et al., 1991) • Vertical jump test corresponding to Abalakow’s vertical jump test (Klavora, 2000) • Balancing backwards from Köperkoordinationstest für Kinder (Kiphard & Schilling, 1974) • Precision throw from “Der Allgemeiner Sportmotorischer Test für Kinder von 6–11 Jahren” (Bös, 2000) • The Andersen test, developed as proxy for maximal oxygen uptake (Andersen et al., 2008) These tests represent various proxy components of motor and aerobic performance. Short shuttle run: agility; handgrip strength: upper extremity strength; vertical jump: coordination and explosive power; balancing backwards: dynamic balance; precision throw: hand-eye coordination; and the Andersen test: aerobic performance.
Data quality A standardized and detailed test protocol was designed for the study. In addition, a mandatory 2-day training workshop was carried out before all test occasions to ensure optimal standardization of the measurements of physical fitness. The workshop included testing of children from a school outside the study setting with relevant age groups. The field workers performed the same one to two fitness tests to increase the accuracy of the testing. The assessors were not blinded with respect to school type. Reliability measures were calculated on the study population and raters through intraclass correlation coefficient (ICC) with a random rater/trial design corresponding to equations presented by Weir (2005) adapted from Shrout and Fleiss’s model 2,k (Shrout & Fleiss, 1979) and McGraw and Wong’s model A,k (McGraw & Wong, 1996). Test–retest ICC for the five motor performance tests ranged between 0.81 and 0.98. Inter-rater reliability ICC was conducted for balancing backward (0.98) and precision throw (0.97). Raw standard error of the mean expressed as percentage of mean for test–retest was 2.5% for short shuttle run test, 6.6% for hand grip strength, 7.6% for vertical jump, 11.4% for balancing backwards, and 28.4% for precision throw. Reliability measures
708
for the Andersen test among schoolchildren as a pilot study for the CHAMPS study DK are described by Ahler et al. (2012).
Other covariates Data at individual and grade level were collected at baseline, e.g., school type (normal PE or extra PE) and gender. Grade level at baseline (preschool to fourth grade) was denoted as cohort followed over time for 3 years. Cohort 0 is children included at grade level 0 followed through grade level 1 and 2. Cohort 1 is children included at grade level 1 and followed trough grade level 2 and 3, etc.
Outcome measure Z-scores were computed for the five motor and one aerobic performance tests using the study population mean and standard derivation for each test at baseline. Thereby, a developmental performance score was created for each child per test occasion relative to baseline values. The individual z-scores were summarized to a composite fitness z-score per test occasion for each child (Brage et al., 2004; Andersen et al., 2006). Children in the lowest 50th percentile of the composite fitness score at baseline, within each cohort, were defined as children with low level of fitness.
Statistical analysis Baseline descriptive statistics were calculated for children with complete baseline data. Simple univariable associations between school type and baseline characteristics were analyzed and tested using a chi-square test (categorical baseline characteristics) and a t-test (continuous baseline characteristics). Multilevel mixedeffects linear regression models were used to examine the association between school type and the longitudinal change in fitness using Stata® (version 11.1 and 12.1, StataCorp, College Station, Texas, USA). Multilevel analyses are suitable for repeated measures over time and for data with a hierarchical structure (Rabe-Hesketh & Skrondal, 2008). Initially, a four-level model was fitted for both research questions with repeated measures (level 1) nested within children (level 2), nested within classes (level 3), and nested within schools (level 4). The estimate of the variation between classes within schools was very small for the model with all children, leading to the use of a three-level model with one variance parameter for school classes combined. Among children within the lowest 50th percentile of fitness at baseline, school level was preferred over classes (the estimate of the variation between classes was very small). The models were constructed with random intercepts of school (low fitness) or school classes (all children), and child level, handling the hierarchical structure and repeated measurements within children. A random slope was included at child level allowing individual variation in the development in fitness. An unstructured covariance structure for children intercept and slope, and maximum likelihood estimation were used for all models. Covariates included in the fixed part of the two models were test occasion, gender, cohort and two interaction terms, a three-way interaction between school type, test occasions, and cohort and a two-way interaction between school type and gender. Inclusion of interaction terms allowed examination of school type differences in the development in fitness across cohorts, and examination of different school type effects in boys and girls, respectively. Significance of fixed effects was evaluated using likelihood ratio test and a 5% significance level. Model fit was investigated graphically at all levels of the models, including individual fitted vs observed values, revealing an overall good model fit with exception of the lowest 15% in a normal quantile plot of all fitted values. A sensitivity analysis excluding the observations corresponding to the lowest 15% residuals resulted in no
Effects of extra physical education in school influential difference in parameter estimates or significance levels. Pre-analysis, school type differences at baseline were investigated using the described model above without time-dependent variables. School type difference in the development in height was also tested pre-analysis as a rough proxy variable for growth and maturation. There was no indication of school type difference in the development of height overall or across cohorts.
Results Study population Eight children with medical conditions affecting motor performance, e.g., spastic hemiplegia, cerebral palsy, or special syndromes, were excluded. Complete fitness data for all five test occasions were available for 686 children. An additional 339 children had complete data at four occasions, 110 children had complete data at three occasions, and 159 children had complete data at one or two test occasions. The high number of drop-in prior to test occasion 6 (Fig. 1) led to the exclusion of 47 late included children ensuring at least 12-month participation prior to the end point of the analyses. Other excluded individuals were children with no complete data on fitness at any test occasion (26 children). Baseline characteristics are shown in Table 1. Raw data investigation indicated a linear development of fitness over time with differences in development across cohorts (Fig. 2). The distribution of children with the lowest fitness at baseline was similar for the two school types: 253 out of 488 children at normal PE schools (52%) and 302 out of 625 children at extra PE schools (48%). There was no difference between school types in the composite z-score for fitness at baseline, both among all children and children in the low fitness at baseline group. Comparison of composite fitness z-score – all children A significant three-way interaction between school type, test occasion, and cohort was found (P = 0.02). No interaction between school type and gender was found. Variance components were estimated to 0.21 standard deviation (SD) for classes, 2.40 SD for children within classes, and 0.22 SD for repeated measurements within children. Borderline significant difference of development in fitness was found for cohort 3. Significant difference was found for cohort 4. Both favoring extra PE over normal PE, corresponding to children entering the project at grade levels 3 and 4 (mean age 9.3 and 10.3 years). The development in fitness (adjusted for differences in gender), school differences, and P-values are shown in Table 2 and illustrated in Fig. 3. Comparison of composite fitness z-score among children with low fitness at baseline The three-way interaction between school type, test occasion, and cohort was not significant, indicating no
variation in school type difference across cohorts. Twoway interactions between school type and test occasion along with test occasion and cohort were both significant. Again, no interaction between school type and gender was found. Variance components were estimated to 0.15 SD for school, 1.27 SD for children within schools, and 0.19 SD for repeated measurements within children. The significant interaction between school type and test occasion indicated a difference in the development in fitness across the two school types among children in the lowest 50th percentile of fitness at baseline. The difference in fitness development between school types was significant (P = 0.018) with a difference in slope of 0.09 composite z-score units per half year [95% confidence interval (CI): 0.02–0.17] favoring extra PE schools. An end point (total) difference of 0.47 z-score units (95% CI: 0.08–0.85) between school types adjusted for the effect of gender and cohort (Fig. 4). Discussion Our study found an overall improved development in a composite fitness score for children enrolled in schools with extra PE entering at grade level four (1.06 z-score units; 95% CI: 0.48–1.65) and a borderline significant difference at grade level three (0.52 z-score units; 95% CI: −0.06–1.09) compared with children in normal PE schools. In the younger grade levels, however, there was no effect of school type on the development of composite fitness score over time. It can be hypothesized that children at younger ages are more physically active during free play activities, thereby not affected in the same manner from the amount of organized PE at school. A recent Australian study (Ridgers et al., 2012) has suggested that daily moderate-to-vigorous physical activity (MVPA) and MVPA performed during school breaks decline with age supporting the hypothesis above. Among the children within the lowest 50th percentile of the composite fitness score at baseline, a school type effect favoring extra PE with 0.47 z-score units (95% CI: 0.08–0.85) was evident after 2.5 years, adjusted for the effect of gender and cohort. Compared with the effect among children with all levels of fitness, this effect is considerable, and dissimilar from the model of children with all levels of fitness where a dependence of cohort is evident in the school type difference (significant threeway interaction between school type, time, and cohort), indicating a positive effect across all cohorts for children with low levels of fitness (non-significant three-way interaction). The Sogndal school intervention study also found an effect of increased PE lessons in school among children with low initial fitness values (Resaland et al., 2011). However, the study only focused on cardiorespiratory fitness. Several studies have found an association between childhood physical fitness or motor
709
Rexen et al. Table 1. Descriptive baseline measures
Baseline characteristics
Normal PE schools
Extra PE schools
Simple tests statistics†
Children Age (years) Gender Male Female Cohorts 0 1st 2nd 3rd 4th Short shuttle run (s) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Hand grip strength (kg) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Vertical jump (cm) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Balancing backward (steps) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Precision throw (points) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Andersen test (m) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Composite fitness z-score Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4
488 8.4 (1.46)
625 8.4 (1.40)
– P = 0.27 P = 0.073
244 244
283 342
82 (46%)‡ 95 (42%)‡ 106 (64%)‡ 98 (48%)‡ 107 (48%)‡
110 (42%)‡ 134 (46%)‡ 135 (47%)‡ 125 (45%)‡ 121 (46%)‡
P = 0.58
29.7 (4.02) 27.0 (3.12) 24.7 (2.35) 24.2 (2.24) 23.6 (2.06)
28.9 (3.65) 26.8 (2.47) 24.9 (2.69) 23.6 (2.07) 23.5 (2.29)
P = 0.12 P = 0.59 P = 0.66 P = 0.051 P = 0.53
12.3 (2.44) 14.5 (2.51) 17.0 (3.49) 17.5 (3.49) 19.7 (4.44)
11.7 (2.46) 13.2 (2.40) 15.3 (3.00) 16.9 (3.26) 18.9 (3.65)
P = 0.24 P < 0.001* P < 0.001* P = 0.25 P = 0.14
19.8 (4.73) 23.6 (4.75) 27.5 (5.71) 30.8 (4.37) 32.3 (6.04)
22.3 (4.70) 24.8 (5.10) 28.4 (5.37) 30.7 (4.94) 31.8 (6.14)
P < 0.001* P = 0.09 P = 0.18 P = 0.54 P = 0.84
28.7 (11.6) 36.1 (12.53) 39.4 (12.73) 46.5 (11.43) 47.7 (11.82)
29.9 (10.20) 33.6 (11.41) 40.0 (11.78) 45.0 (12.18) 49.2 (11.60)
P = 0.43 P = 0.10 P = 0.69 P = 0.27 P = 0.29
5.9 (3.88) 8.4 (4.31) 11.2 (3.56) 12.4 (3.84) 14.5 (4.05)
6.0 (3.80) 8.4 (4.03) 11.9 (4.24) 14.0 (4.31) 15.1 (3.66)
P = 0.87 P = 0.79 P = 0.21 P = 0.005* P = 0.24
796 (91.4) 860 (90.8) 907 (105.3) 931 (103.4) 936 (100.5)
827 (72.0) 865 (69.5) 904 (105.0) 942 (101.0) 940 (101.4)
P = 0.010* P = 0.61 P = 0.83 P = 0.43 P = 0.72
−5.74 (2.69) −2.20 (2.74) 0.80 (2.95) 2.56 (2.96) 4.04 (2.75)
−4.43 (2.70) −2.43 (2.45) 0.66 (3.20) 2.87 (3.09) 4.08 (2.94)
P = 0.026* P = 0.51 P = 0.72 P = 0.45 P = 0.92
Components of the composite physical fitness score and information by school type and cohort. Age and fitness measures expressed as mean (standard deviation). † Chi-square test (gender, cohort) and t-test (age, fitness measures), *significant difference. ‡ Gender distribution within each cohort by school type (normal PE and extra PE) given in % boys. PE, physical education.
skills and adolescent fitness levels (Malina, 1996; Twisk et al., 2000; Matton et al., 2006; Barnett et al., 2008). Considering this phenomenon, referred to as tracking, improved childhood fitness has the possibility to improve fitness levels in adolescence, thereby potentially
710
combating various health-related risk factors in children with low levels of fitness (Brage et al., 2004; Andersen et al., 2006). This study uses a composite physical fitness score, limiting the possibility of comparison within similar
Effects of extra physical education in school
Fig. 2. Development in raw composite physical fitness z-score over time by school type [normal or extra physical education (PE) school] across cohort. Cohort 0 is children included at grade level 0 followed through grade level 1 and 2. Cohort 1 is children included at grade level 1 and followed trough grade level 2 and 3, etc.
populations. However, the effects of development in fitness from schools with extra PE are universally applicable in most settings. Kriemler et al. (2011) stated that school-based PA interventions have a potential public health effect of increasing PA and possibly also aerobic
fitness in healthy children and adolescents. However, the review (Kriemler et al., 2011) showed inconclusive findings on the effect of motor skills, mostly due to the diversity in measurements including mixture of qualitative and quantitative scoring.
711
Rexen et al. Table 2. Estimated total development of composite physical fitness across cohort by school type
Normal PE 95% CI Extra PE 95% CI Difference in fitness slopes P-value
Cohort 0
Cohort 1
Cohort 2
Cohort 3
Cohort 4
7.16 6.69 to 7.64 6.83 6.43 to 7.23 −0.33 −0.95 to 0.29 0.29
5.84 5.40 to 6.29 6.30 5.93 to 6.68 0.46 −0.12 to 1.04 0.12
4.74 4.31 to 5.16 4.84 4.46 to 5.21 0.10 −0.47 to 0.66 0.73
4.59 4.16 to 5.02 5.11 4.72 to 5.49 0.52 −0.06 to 1.09 0.078
4.62 4.21 to 5.04 5.69 5.28 to 6.09 1.06 0.48 to 1.65
© 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Effects of extra school-based physical education on overall physical fitness development – the CHAMPS study DK C. T. Rexen1, A. K. Ersbøll2, N. C. Møller1, H. Klakk1,3 N. Wedderkopp1,4,5, L. B. Andersen1,6 1
Centre of Research in Childhood Health, Institute of Sports Science and Clinical Biomechanics, University of Southern Denmark, Odense M, Denmark, 2National Institute of Public Health, University of Southern Denmark, København K, Denmark, 3University College Lillebaelt, Odense M, Denmark, 4Institute of Regional Health Services Research, University of Southern Denmark, Odense M, Denmark, 5Sport Medicine Clinic, Orthopedic Department, Hospital of Lillebaelt, Middelfart, Denmark, 6Department of Sports Medicine, The Norwegian School of Sport Sciences, Oslo, Norway Corresponding author: Christina Trifonov Rexen, Centre of Research in Childhood Health, Institute of Sports Science and Clinical Biomechanics, University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark. Tel: +45 2671 9691, E-mail: [email protected] Accepted for publication 19 June 2014
First, this study aimed to investigate if four extra physical education (PE) lessons per week improved children’s development in physical fitness. Second, to investigate if the extra PE lessons improved development in physical fitness for children with lower levels of fitness at baseline. This study was a longitudinal controlled school-based study. The study population consisted of 10 Danish public schools with children in preschool to fourth grade (cohorts 0–4) with 2.5-year follow-up. Six schools had extra PE and four schools had normal PE. In total 1247 children were included (normal PE = 536, extra PE = 711). Development in fitness was analyzed using a composite z-score from six fitness tests. Multilevel mixed-
effects linear regression was used to examine the association between school type and development in fitness. Extra PE increased the total development of composite z-score units among children enrolled in cohort 4 and borderline in cohort 3 with 1.06 (95% confidence interval 0.48–1.65) and 0.52 z-score units (−0.06 to 1.09), respectively. Children in the lower 50 percentiles increased their development with 0.47 (0.08–0.85) z-score units. Extra PE in schools improved development in fitness for cohort 4 and borderline for cohort 3 among all children. Extra PE improved fitness development across all cohorts among children with low fitness levels.
A polarization in fitness and other health-related measures among school-aged youth has been identified over the past decades (Wedderkopp et al., 2004; Dyrstad et al., 2012). It is not yet clear whether this can be restored through school-based interventions. Furthermore, the optimal quality, quantity, and intensity in school settings have yet to be discovered, although multicomponent interventions have already been shown to have a positive effect on the overall level of physical activity (PA) and possibly also a positive effect on fitness among school-aged youth (Kriemler et al., 2011). School-based interventions evaluating physical fitness components have identified positive effects on cardiorespiratory fitness (Hansen et al., 1991; Resaland et al., 2011) and several physical fitness components (Sollerhed & Ejlertsson, 2008). However, school interventions with no significant effects on physical fitness have also been reported (Bugge et al., 2012; Magnusson et al., 2012; Christiansen et al., 2013). A review by Kriemler et al. (2011) found positive effects of schoolbased interventions on fitness in 6 out of 11 studies,
suggesting that school setting could be an important arena for restoration of the current negative trend in fitness among healthy youth. The question of how much extra physical education (PE) is needed to document longitudinal effects on fitness is not easily answered. A review by Strong et al. (2005) suggests at least 60 min of moderate-to-vigorous daily PA for school-aged youth to accomplish effects on aerobic fitness, muscular strength and endurance, and other health-related effects of regular PA. Motor performance or fundamental movement skills are considered the foundation of more complex and specialized movement sequences required for participation in organized sport as well as many non-organized activities during free play in childhood and adolescence (Clark & Metcalfe, 2002; Gallahue & Ozmun, 2006). Most movements involved in an active lifestyle have elements of both gross and fine motor components (Payne & Isaacs, 2002). Fundamental movement skills are developed in childhood and subsequently refined into more complex context- and sport-specific skills (Clark, 2005;
706
Effects of extra physical education in school Stodden et al., 2008) during early and middle school years (Gallahue & Ozmun, 2006). Therefore, it seems relevant to combine testing of motor performance (quantitatively assessable movement skill; Gallahue & Ozmun, 2006) with measures of cardiorespiratory fitness in children when evaluating the effects of extra PE lessons. Furthermore, school-based interventions aiming to increase PA and fitness are thought to be effective and universally applicable because PE in schools is compulsory in most countries and includes the least active children (Kriemler et al., 2011) at a developmentally important age. Children with low levels of fitness are particularly interesting from a health-related perspective (Brage et al., 2004; Andersen et al., 2006). Objectives First, the objective of this study was to investigate if six weekly PE lessons were sufficient to improve physical fitness development in schoolchildren when compared with the traditional two weekly lessons in normal schools. Second, to investigate if children with low levels of fitness at baseline improved their development of physical fitness in schools with extra PE compared with normal schools. Both aims were answered using intention-to-treat analyses. Methods Study population The CHAMPS study DK is a longitudinal study. The initial concept was brought forth at municipality level by local authorities and extended by researchers to include control schools. The study design can therefore be described as a quasi-experimental study evaluating a natural experiment (Craig et al., 2012). In Denmark, children attend primary public schools from the age of 6 (preschool or grade 0) to the age of 16 or 17 (9th or 10th grade). For smaller schools, this may involve school change at grade level 7. All public schools in Denmark are under the influence of governmental reform at a national level. This reform is adapted to match resources and priorities within each municipality by local authorities. Finally, detailed planning is made at school level by the school itself. In the CHAMPS study DK, all 19 primary schools in the municipality of Svendborg (Southern Funen, Denmark) were invited to participate in the project as schools with extra PE (270 min/week divided over at least three sessions per week). Ten schools agreed to participate, but only six schools were willing to prioritize the financing of the extra PE sessions. To study the effect of extra PE in schools, the municipality was asked to obtain matched control schools according to school size, urban/rural area, and socioeconomic position among the 13 schools that continued the normal curriculum of 90-min PE per week (one session). This resulted in a total of n = 773 and n = 734 invited children in schools with extra PE and normal PE, respectively. Parents and children were kept unaware of the initiation of this project until after the deadline for school choice, thereby avoiding influence by parents on the choice of school. All 10 schools in the study accommodated all grade levels within the study period (0–7th grade). Detailed description of the study protocol is published elsewhere (Wedderkopp et al., 2012). The children were followed for 3 years. At baseline (August 2008) 1222 children
Fig. 1. Flow chart of participant at various time points. The number and school type [normal or extra physical education (PE) school] of drop-in or out are shown. 1 Eighty-one percent of the total invited participants agreed to participate (at test occasion 1). 2In total, 71% of the invited participants at schools with normal PE agreed to participate. Ninety-one percent of the children at extra PE schools agreed to participate. From the total of 99 dropouts, 83 were due to change of school. participated (by school type: normal PE: n = 522, extra PE: n = 700). The project was continuously open for new participants, and a total of 99 dropout and 108 drop-in were observed during the 2.5-year follow-up (Fig. 1).
Socioeconomic status There were no significant differences in household income between the two school types. However, household income was 15% lower in non-participating schools. There was no significant difference of parental education between school types or between participating vs non-participating schools in the municipality.
Ethics The study was carried out in accordance with the Declaration of Helsinki and was approved by the local scientific ethics committee (ID S20080047) and registered in the Danish Data Protection Agency (J.nr. 2008-41-2240). Parents of the children gave written informed consent and the children gave verbal consent.
PE lessons in schools with extra PE To ensure optimal quality of the PE lessons, PE teachers and pedagogues participated in a 40-lesson course in “Age-related concepts of training” organized by Team Denmark in collaboration with the faculty of education at the University College Lillebaelt December 2008–April 2009. The goal was to
707
Rexen et al. ensure targeted and responsible training for the growing individual (Pryce et al., 2005). Furthermore, a handbook was provided by Team Denmark with inspiration for the content of PE lessons, practical suggestions, and exercises. The handbook is published by Bach and Eiberg (2010). Overall, the concept involves focus on gradual development of all-round fundamental motor skills for the prepubertal children through organized playing activities involving training of movement skills, e.g., running, hopping and jumping, throwing and catching, kick and dribble, orientation skills and sensation of positions, simple balance, rhythm, coordination, reaction, and timing. As the children enter puberty, the activities and training becomes more complex and advanced and approaches more sports-specific skills, e.g., activities with repeated hops and jumps, change of direction, speed, and more advanced ball games including ball-oriented timing and agility.
Physical fitness testing Participating children were tested in total five times during a period of three school years. Four tests were performed with half a year interval during the first 2 years and again 1 year later at the end of the 3-year period. Baseline data were collected in September–October 2008. The quantitative fitness testing consisted of five motor performance tests and one aerobic performance test. These tests were: • Short shuttle run test from the Eurofit test battery (Adam et al., 1991) • Handgrip strength test from the Eurofit test battery (Adam et al., 1991) • Vertical jump test corresponding to Abalakow’s vertical jump test (Klavora, 2000) • Balancing backwards from Köperkoordinationstest für Kinder (Kiphard & Schilling, 1974) • Precision throw from “Der Allgemeiner Sportmotorischer Test für Kinder von 6–11 Jahren” (Bös, 2000) • The Andersen test, developed as proxy for maximal oxygen uptake (Andersen et al., 2008) These tests represent various proxy components of motor and aerobic performance. Short shuttle run: agility; handgrip strength: upper extremity strength; vertical jump: coordination and explosive power; balancing backwards: dynamic balance; precision throw: hand-eye coordination; and the Andersen test: aerobic performance.
Data quality A standardized and detailed test protocol was designed for the study. In addition, a mandatory 2-day training workshop was carried out before all test occasions to ensure optimal standardization of the measurements of physical fitness. The workshop included testing of children from a school outside the study setting with relevant age groups. The field workers performed the same one to two fitness tests to increase the accuracy of the testing. The assessors were not blinded with respect to school type. Reliability measures were calculated on the study population and raters through intraclass correlation coefficient (ICC) with a random rater/trial design corresponding to equations presented by Weir (2005) adapted from Shrout and Fleiss’s model 2,k (Shrout & Fleiss, 1979) and McGraw and Wong’s model A,k (McGraw & Wong, 1996). Test–retest ICC for the five motor performance tests ranged between 0.81 and 0.98. Inter-rater reliability ICC was conducted for balancing backward (0.98) and precision throw (0.97). Raw standard error of the mean expressed as percentage of mean for test–retest was 2.5% for short shuttle run test, 6.6% for hand grip strength, 7.6% for vertical jump, 11.4% for balancing backwards, and 28.4% for precision throw. Reliability measures
708
for the Andersen test among schoolchildren as a pilot study for the CHAMPS study DK are described by Ahler et al. (2012).
Other covariates Data at individual and grade level were collected at baseline, e.g., school type (normal PE or extra PE) and gender. Grade level at baseline (preschool to fourth grade) was denoted as cohort followed over time for 3 years. Cohort 0 is children included at grade level 0 followed through grade level 1 and 2. Cohort 1 is children included at grade level 1 and followed trough grade level 2 and 3, etc.
Outcome measure Z-scores were computed for the five motor and one aerobic performance tests using the study population mean and standard derivation for each test at baseline. Thereby, a developmental performance score was created for each child per test occasion relative to baseline values. The individual z-scores were summarized to a composite fitness z-score per test occasion for each child (Brage et al., 2004; Andersen et al., 2006). Children in the lowest 50th percentile of the composite fitness score at baseline, within each cohort, were defined as children with low level of fitness.
Statistical analysis Baseline descriptive statistics were calculated for children with complete baseline data. Simple univariable associations between school type and baseline characteristics were analyzed and tested using a chi-square test (categorical baseline characteristics) and a t-test (continuous baseline characteristics). Multilevel mixedeffects linear regression models were used to examine the association between school type and the longitudinal change in fitness using Stata® (version 11.1 and 12.1, StataCorp, College Station, Texas, USA). Multilevel analyses are suitable for repeated measures over time and for data with a hierarchical structure (Rabe-Hesketh & Skrondal, 2008). Initially, a four-level model was fitted for both research questions with repeated measures (level 1) nested within children (level 2), nested within classes (level 3), and nested within schools (level 4). The estimate of the variation between classes within schools was very small for the model with all children, leading to the use of a three-level model with one variance parameter for school classes combined. Among children within the lowest 50th percentile of fitness at baseline, school level was preferred over classes (the estimate of the variation between classes was very small). The models were constructed with random intercepts of school (low fitness) or school classes (all children), and child level, handling the hierarchical structure and repeated measurements within children. A random slope was included at child level allowing individual variation in the development in fitness. An unstructured covariance structure for children intercept and slope, and maximum likelihood estimation were used for all models. Covariates included in the fixed part of the two models were test occasion, gender, cohort and two interaction terms, a three-way interaction between school type, test occasions, and cohort and a two-way interaction between school type and gender. Inclusion of interaction terms allowed examination of school type differences in the development in fitness across cohorts, and examination of different school type effects in boys and girls, respectively. Significance of fixed effects was evaluated using likelihood ratio test and a 5% significance level. Model fit was investigated graphically at all levels of the models, including individual fitted vs observed values, revealing an overall good model fit with exception of the lowest 15% in a normal quantile plot of all fitted values. A sensitivity analysis excluding the observations corresponding to the lowest 15% residuals resulted in no
Effects of extra physical education in school influential difference in parameter estimates or significance levels. Pre-analysis, school type differences at baseline were investigated using the described model above without time-dependent variables. School type difference in the development in height was also tested pre-analysis as a rough proxy variable for growth and maturation. There was no indication of school type difference in the development of height overall or across cohorts.
Results Study population Eight children with medical conditions affecting motor performance, e.g., spastic hemiplegia, cerebral palsy, or special syndromes, were excluded. Complete fitness data for all five test occasions were available for 686 children. An additional 339 children had complete data at four occasions, 110 children had complete data at three occasions, and 159 children had complete data at one or two test occasions. The high number of drop-in prior to test occasion 6 (Fig. 1) led to the exclusion of 47 late included children ensuring at least 12-month participation prior to the end point of the analyses. Other excluded individuals were children with no complete data on fitness at any test occasion (26 children). Baseline characteristics are shown in Table 1. Raw data investigation indicated a linear development of fitness over time with differences in development across cohorts (Fig. 2). The distribution of children with the lowest fitness at baseline was similar for the two school types: 253 out of 488 children at normal PE schools (52%) and 302 out of 625 children at extra PE schools (48%). There was no difference between school types in the composite z-score for fitness at baseline, both among all children and children in the low fitness at baseline group. Comparison of composite fitness z-score – all children A significant three-way interaction between school type, test occasion, and cohort was found (P = 0.02). No interaction between school type and gender was found. Variance components were estimated to 0.21 standard deviation (SD) for classes, 2.40 SD for children within classes, and 0.22 SD for repeated measurements within children. Borderline significant difference of development in fitness was found for cohort 3. Significant difference was found for cohort 4. Both favoring extra PE over normal PE, corresponding to children entering the project at grade levels 3 and 4 (mean age 9.3 and 10.3 years). The development in fitness (adjusted for differences in gender), school differences, and P-values are shown in Table 2 and illustrated in Fig. 3. Comparison of composite fitness z-score among children with low fitness at baseline The three-way interaction between school type, test occasion, and cohort was not significant, indicating no
variation in school type difference across cohorts. Twoway interactions between school type and test occasion along with test occasion and cohort were both significant. Again, no interaction between school type and gender was found. Variance components were estimated to 0.15 SD for school, 1.27 SD for children within schools, and 0.19 SD for repeated measurements within children. The significant interaction between school type and test occasion indicated a difference in the development in fitness across the two school types among children in the lowest 50th percentile of fitness at baseline. The difference in fitness development between school types was significant (P = 0.018) with a difference in slope of 0.09 composite z-score units per half year [95% confidence interval (CI): 0.02–0.17] favoring extra PE schools. An end point (total) difference of 0.47 z-score units (95% CI: 0.08–0.85) between school types adjusted for the effect of gender and cohort (Fig. 4). Discussion Our study found an overall improved development in a composite fitness score for children enrolled in schools with extra PE entering at grade level four (1.06 z-score units; 95% CI: 0.48–1.65) and a borderline significant difference at grade level three (0.52 z-score units; 95% CI: −0.06–1.09) compared with children in normal PE schools. In the younger grade levels, however, there was no effect of school type on the development of composite fitness score over time. It can be hypothesized that children at younger ages are more physically active during free play activities, thereby not affected in the same manner from the amount of organized PE at school. A recent Australian study (Ridgers et al., 2012) has suggested that daily moderate-to-vigorous physical activity (MVPA) and MVPA performed during school breaks decline with age supporting the hypothesis above. Among the children within the lowest 50th percentile of the composite fitness score at baseline, a school type effect favoring extra PE with 0.47 z-score units (95% CI: 0.08–0.85) was evident after 2.5 years, adjusted for the effect of gender and cohort. Compared with the effect among children with all levels of fitness, this effect is considerable, and dissimilar from the model of children with all levels of fitness where a dependence of cohort is evident in the school type difference (significant threeway interaction between school type, time, and cohort), indicating a positive effect across all cohorts for children with low levels of fitness (non-significant three-way interaction). The Sogndal school intervention study also found an effect of increased PE lessons in school among children with low initial fitness values (Resaland et al., 2011). However, the study only focused on cardiorespiratory fitness. Several studies have found an association between childhood physical fitness or motor
709
Rexen et al. Table 1. Descriptive baseline measures
Baseline characteristics
Normal PE schools
Extra PE schools
Simple tests statistics†
Children Age (years) Gender Male Female Cohorts 0 1st 2nd 3rd 4th Short shuttle run (s) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Hand grip strength (kg) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Vertical jump (cm) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Balancing backward (steps) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Precision throw (points) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Andersen test (m) Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4 Composite fitness z-score Cohort 0 Cohort 1 Cohort 2 Cohort 3 Cohort 4
488 8.4 (1.46)
625 8.4 (1.40)
– P = 0.27 P = 0.073
244 244
283 342
82 (46%)‡ 95 (42%)‡ 106 (64%)‡ 98 (48%)‡ 107 (48%)‡
110 (42%)‡ 134 (46%)‡ 135 (47%)‡ 125 (45%)‡ 121 (46%)‡
P = 0.58
29.7 (4.02) 27.0 (3.12) 24.7 (2.35) 24.2 (2.24) 23.6 (2.06)
28.9 (3.65) 26.8 (2.47) 24.9 (2.69) 23.6 (2.07) 23.5 (2.29)
P = 0.12 P = 0.59 P = 0.66 P = 0.051 P = 0.53
12.3 (2.44) 14.5 (2.51) 17.0 (3.49) 17.5 (3.49) 19.7 (4.44)
11.7 (2.46) 13.2 (2.40) 15.3 (3.00) 16.9 (3.26) 18.9 (3.65)
P = 0.24 P < 0.001* P < 0.001* P = 0.25 P = 0.14
19.8 (4.73) 23.6 (4.75) 27.5 (5.71) 30.8 (4.37) 32.3 (6.04)
22.3 (4.70) 24.8 (5.10) 28.4 (5.37) 30.7 (4.94) 31.8 (6.14)
P < 0.001* P = 0.09 P = 0.18 P = 0.54 P = 0.84
28.7 (11.6) 36.1 (12.53) 39.4 (12.73) 46.5 (11.43) 47.7 (11.82)
29.9 (10.20) 33.6 (11.41) 40.0 (11.78) 45.0 (12.18) 49.2 (11.60)
P = 0.43 P = 0.10 P = 0.69 P = 0.27 P = 0.29
5.9 (3.88) 8.4 (4.31) 11.2 (3.56) 12.4 (3.84) 14.5 (4.05)
6.0 (3.80) 8.4 (4.03) 11.9 (4.24) 14.0 (4.31) 15.1 (3.66)
P = 0.87 P = 0.79 P = 0.21 P = 0.005* P = 0.24
796 (91.4) 860 (90.8) 907 (105.3) 931 (103.4) 936 (100.5)
827 (72.0) 865 (69.5) 904 (105.0) 942 (101.0) 940 (101.4)
P = 0.010* P = 0.61 P = 0.83 P = 0.43 P = 0.72
−5.74 (2.69) −2.20 (2.74) 0.80 (2.95) 2.56 (2.96) 4.04 (2.75)
−4.43 (2.70) −2.43 (2.45) 0.66 (3.20) 2.87 (3.09) 4.08 (2.94)
P = 0.026* P = 0.51 P = 0.72 P = 0.45 P = 0.92
Components of the composite physical fitness score and information by school type and cohort. Age and fitness measures expressed as mean (standard deviation). † Chi-square test (gender, cohort) and t-test (age, fitness measures), *significant difference. ‡ Gender distribution within each cohort by school type (normal PE and extra PE) given in % boys. PE, physical education.
skills and adolescent fitness levels (Malina, 1996; Twisk et al., 2000; Matton et al., 2006; Barnett et al., 2008). Considering this phenomenon, referred to as tracking, improved childhood fitness has the possibility to improve fitness levels in adolescence, thereby potentially
710
combating various health-related risk factors in children with low levels of fitness (Brage et al., 2004; Andersen et al., 2006). This study uses a composite physical fitness score, limiting the possibility of comparison within similar
Effects of extra physical education in school
Fig. 2. Development in raw composite physical fitness z-score over time by school type [normal or extra physical education (PE) school] across cohort. Cohort 0 is children included at grade level 0 followed through grade level 1 and 2. Cohort 1 is children included at grade level 1 and followed trough grade level 2 and 3, etc.
populations. However, the effects of development in fitness from schools with extra PE are universally applicable in most settings. Kriemler et al. (2011) stated that school-based PA interventions have a potential public health effect of increasing PA and possibly also aerobic
fitness in healthy children and adolescents. However, the review (Kriemler et al., 2011) showed inconclusive findings on the effect of motor skills, mostly due to the diversity in measurements including mixture of qualitative and quantitative scoring.
711
Rexen et al. Table 2. Estimated total development of composite physical fitness across cohort by school type
Normal PE 95% CI Extra PE 95% CI Difference in fitness slopes P-value
Cohort 0
Cohort 1
Cohort 2
Cohort 3
Cohort 4
7.16 6.69 to 7.64 6.83 6.43 to 7.23 −0.33 −0.95 to 0.29 0.29
5.84 5.40 to 6.29 6.30 5.93 to 6.68 0.46 −0.12 to 1.04 0.12
4.74 4.31 to 5.16 4.84 4.46 to 5.21 0.10 −0.47 to 0.66 0.73
4.59 4.16 to 5.02 5.11 4.72 to 5.49 0.52 −0.06 to 1.09 0.078
4.62 4.21 to 5.04 5.69 5.28 to 6.09 1.06 0.48 to 1.65
