Scaling up Action Schools! BC: How Does Voltage Drop at Scale Affect Student Level Outcomes? A Cluster Randomized Controlled Trial

Action Schools! BC (AS! BC) was scaled-up from an efficacy trial to province-wide delivery across 11 years (2004–2015). In this study we: (1) describe strategies that supported implementation and scale-up; (2) evaluate implementation (teachers’ physical activity (PA) delivery) and student’s PA and cardiorespiratory fitness (CRF) within a cluster randomized controlled trial during years 2 and 3 of scale-up; and (3) assess relationships between teacher-level implementation and student-level outcomes. We classified implementation strategies as process, capacity-building or scale-up strategies. Elementary schools (n = 30) were randomized to intervention (INT; 16 schools; 747 students) or usual practice (UP; 14 schools; 782 students). We measured teachers’ PA delivery (n = 179) using weekly logs; students’ PA by questionnaire (n = 30 schools) and accelerometry (n = 9 schools); and students’ CRF by 20-m shuttle run (n = 25 schools). INT teachers delivered more PA than UP teachers in year 1 (+33.8 min/week, 95% CI 12.7, 54.9) but not year 2 (+18.8 min/week, 95% CI −0.8, 38.3). Unadjusted change in CRF was 36% and 27% higher in INT girls and boys, respectively, compared with their UP peers (year 1; effect size 0.28–0.48). Total PA delivered was associated with change in children’s self-reported MVPA (year 1; r = 0.17, p = 0.02). Despite the ‘voltage drop’, scaling-up school-based PA models is feasible and may enhance children’s health. Stakeholders must conceive of new ways to effectively sustain scaled-up health promoting interventions if we are to improve the health of students at a population level. Clinical Trials registration: NCT01412203.


Introduction
Physical activity (PA) [1] and cardiorespiratory fitness (CRF) [2] are powerful, independent [3] indicators of child and youth health. Yet, relatively few children and youth engage in recommended amounts of PA [4]; national and international data show secular declines in both PA [5,6] and CRF [6][7][8]. PA and CRF track across childhood [9], into adolescence [10] and early adulthood [11,12]. High CRF in adolescence is associated with reduced risk of cardiovascular events, cancer and premature mortality in adulthood [13][14][15]. Therefore, effective strategies that reach large numbers of young children are urgently needed to promote PA and CRF, and to maintain behaviors across the life course.
Schools are a focal setting for health promotion [16]; 'whole-of-school' models are one of eight best investments to enhance PA [17]. Children spend nearly half their waking hours at school, and schools reach diverse populations of children across sociodemographic and socioeconomic backgrounds [18]. However, most interventions do not extend beyond short-term (e.g., one month to one school year in length) well-controlled efficacy Table 1. Key components of the Action Schools! BC (AS! BC) model. Reprinted from Journal of Science and Medicine in Sport 9(5), Naylor et al [28], Lessons learned from Action Schools! BC-An 'active school' model to promote physical activity in elementary schools, p413-423 (2006), with permission from Elsevier.

Component Description
Action Zones Six areas in which opportunities for physical activity could be provided to students. The six Zones were: (1) School Environment, (2) School Spirit, (3) Physical Education, (4) Extra-curricular, (5) Family and Community and (6) Classroom Action.

AS! BC Support Team
A central technical support unit that developed and provided AS! BC resources (training workshops, written materials, Classroom Action Bins, school newsletter inserts for families) and ongoing consultation (on-site and telephone) to administrators, teachers and the School Action Team.

AS! BC School Facilitators
Two elementary school teachers seconded by the AS! BC Support Team to provide training, support and advice to the schools and liaise between the Support Team and the School Action Team.

School Action Team
A committee of school stakeholders (e.g., interested intermediate grade teachers, administrators, parents, health, sport/recreation practitioners) that created and supported implementation of the Action Plan.
Planning Guide for Schools and Teachers A set of inventories and worksheets that guided teachers and the School Action Team to identify school priorities and create their Action Plan.
Action Pages! A resource directory using curriculum organizers to link teachers, coaches or community instructors with recommended and available resources.

Classroom Action Bin
A storage bin for the classroom filled with playground balls, videos, skipping ropes, exercise bands, strength grippers and teaching resources that supported the Action Plan.
As per models that describe 'essential' elements of implementation and scale-up [36,37], the AS! BC Support Team (JW Sporta; Support Team) was crucial to AS! BC implementation and scale-up success [28]. JW Sporta included experts in healthy living and education who had worked within the BC elementary school system for > 20 years to develop and deliver sport education resources. As the cornerstone of implementation, they established 'resource teams' [37] who guided all aspects of implementation and evaluation of AS! BC in schools province-wide. That is, they worked closely with a host of school community partners at regional and provincial levels to develop AS! BC materials, tools and products, and systematically built capacity within schools to deliver the intervention [28]. The Support Team widely promoted AS! BC, worked to build and maintain partnerships with schools and a broad range of local, provincial, national and international stakeholders, and collaborated with the research team to design and conduct the evaluation. During scale-up, the Support Team was comprised annually of 5-8 full and part-time staff and various contractors who worked with~75 teacher/facilitator trainers annually. Teachers/facilitators who delivered AS! BC workshops had travel and training expenses covered and received either an honorarium or were bought out of teaching time (Teacher on Call) at their own school as needed to deliver training.
We provide a detailed overview of AS! BC's 'comprehensive pathway' [33] to scaleup elsewhere [34] and provide a timeline in Figure 1. With continued support from BC Ministry of Health, we built on the success of AS! BC, demonstrated in our efficacy study [27,29,30], and initiated provincial scale-up in 2004. AS! BC began as a PA initiative for grade 4-7 students; however, the program was continually enhanced to include Healthy Eating (2006 [38]), PA for kindergarten-grade 3 (2006), PA Student Leadership (2006), and Healthy Eating Student Leadership (2011).

Objective 1: Implementation Strategies across 11 Years of Province-Wide Scale up (2004-2015)
To address our first objective, we retrospectively reviewed the Support Team's detailed annual reports and monthly statistical summaries. We first generated a list of implementation strategies, defined as "methods or techniques used to enhance adoption, implementation, and sustainability of evidence-based interventions (EBIs)" [39]. We grouped implementation strategies into three broad categories [40]: (a) implementation process; (b) capacity-building and; (c) scale-up. Implementation process strategies are "processes or activities that implementation or quality improvement (resource) teams perform to plan, select, and integrate an EBI into practice" [40]. Capacity-building strategies targeted individuals' general capacity (motivation, self-efficacy) to execute implementation process strategies (e.g., training and technical assistance) [40]. Scale-up strategies enacted by the support system targeted implementing a specific EBI in multiple settings [40]. We map our strategies to standardized language [39] and follow reporting guidelines [41] to describe each strategy. We extracted data on reach of AS! BC at scale-up (e.g., number of schools registered), and implementation strategies at scale-up (e.g., number of workshops delivered, number of participants attending) from the Support Team's reports.

Objectives 2 & 3: Randomized Control Effectiveness Trial (2005-2007)
We conducted a cluster randomized controlled trial of AS! BC, that incorporated aspects of a hybrid type I effectiveness-implementation study design [42] in a subset of schools (Clinical Trials Registry NCT01412203) beginning in year 2 of provincial scale-up. A school cluster design was used given the whole-school nature of AS! BC. Evaluation took place over two school years (2005)(2006)(2007) with measurement scheduled for the beginning and end of each school year (T1-T4; Figure 1).

Study Design and Participants
The research team introduced AS! BC and the research study at the annual provincewide principals and vice principal's meeting in Vancouver, Canada (2005). A formal letter inviting principals to participate was sent to all BC schools. Administrators from eightyseven schools from across BC expressed interest; we contacted them and 30 schools that met our inclusion criteria (not currently participating in AS! BC or any other PA or healthy  To address our first objective, we retrospectively reviewed the Support Team's detailed annual reports and monthly statistical summaries. We first generated a list of implementation strategies, defined as "methods or techniques used to enhance adoption, implementation, and sustainability of evidence-based interventions (EBIs)" [39]. We grouped implementation strategies into three broad categories [40]: (a) implementation process; (b) capacity-building and; (c) scale-up. Implementation process strategies are "processes or activities that implementation or quality improvement (resource) teams perform to plan, select, and integrate an EBI into practice" [40]. Capacity-building strategies targeted individuals' general capacity (motivation, self-efficacy) to execute implementation process strategies (e.g., training and technical assistance) [40]. Scale-up strategies enacted by the support system targeted implementing a specific EBI in multiple settings [40]. We map our strategies to standardized language [39] and follow reporting guidelines [41] to describe each strategy. We extracted data on reach of AS! BC at scale-up (e.g., number of schools registered), and implementation strategies at scale-up (e.g., number of workshops delivered, number of participants attending) from the Support Team's reports.

Objectives 2 & 3: Randomized Control Effectiveness Trial (2005-2007)
We conducted a cluster randomized controlled trial of AS! BC, that incorporated aspects of a hybrid type I effectiveness-implementation study design [42] in a subset of schools (Clinical Trials Registry NCT01412203) beginning in year 2 of provincial scale-up. A school cluster design was used given the whole-school nature of AS! BC. Evaluation took place over two school years (2005)(2006)(2007) with measurement scheduled for the beginning and end of each school year (T1-T4; Figure 1).

Study Design and Participants
The research team introduced AS! BC and the research study at the annual provincewide principals and vice principal's meeting in Vancouver, Canada (2005). A formal letter inviting principals to participate was sent to all BC schools. Administrators from eightyseven schools from across BC expressed interest; we contacted them and 30 schools that met our inclusion criteria (not currently participating in AS! BC or any other PA or healthy eating program beyond physical education (PE)) volunteered to participate ( Figure 2). We invited all grade 4 and 5 teachers in volunteer schools to participate in the evaluation. Schools were stratified by size (≥300 or <300 students) and geographic location (5 health regions) and randomly assigned to intervention (INT; n = 16 schools) or usual practice (UP; n = 14 schools) conditions by an external research group using a computer generated sequence. Schools randomized to the UP group maintained their regular activities. eating program beyond physical education (PE)) volunteered to participate ( Figure 2). We invited all grade 4 and 5 teachers in volunteer schools to participate in the evaluation. Schools were stratified by size (≥300 or <300 students) and geographic location (5 health regions) and randomly assigned to intervention (INT; n = 16 schools) or usual practice (UP; n = 14 schools) conditions by an external research group using a computer generated sequence. Schools randomized to the UP group maintained their regular activities.  All grade 4 and 5 students (of participating teachers) who participated in regular PE and provided written informed consent from a parent or guardian, were eligible for the evaluation. Schools incorporated AS! BC into regular programming; all students took part in the intervention, regardless of whether they consented to be evaluated. Research ethics boards at the University of British Columbia (B05-0505) and University of Victoria (07-05-149f) approved this investigation.

Implementation Measures (Physical Activity Delivery by Teachers)
All teachers were asked to complete weekly PA logs to record the PA that their class engaged in; this included PE and additional PA such as Classroom Action (INT schools only) and PA-related field trips. For each teacher, we calculated the percentage of weekly logs returned and average weekly PA delivered (min/week). For teachers at INT schools, we also calculated average weekly Classroom Action PA (min/week) as an indicator of fidelity to the model.

Impact Measures (Student-Level Outcomes)
Trained research staff collected all data; it was not possible to blind research staff to group assignment. Children were excused from their classroom in small groups for measurement (~6/group), which took place in school gymnasia, common spaces, or outdoors Self-reported physical activity: We measured moderate-to-vigorous PA (MVPA) over the previous 7 days in the whole cohort (n = 30 schools) using the valid and reliable Physical Activity Questionnaire for Children (PAQ-C) [43], which we modified to include an estimate of time spent in various leisure-time activities in question 1 [44]. Research assistants administered the questionnaire in small groups of 4-6 students. We averaged PAQ-C items to create a general PA score ranging from 1 (low active) to 5 (high active) and obtained an estimate of time (min/day; MVPA PAQ ) spent in MVPA from question 1. Estimated time in MVPA (min/day) was positively skewed; 42 participants (3% of total observations) self-reported MVPA PAQ ≥ 500 min/day. We limited the maximum possible value to 500 min/day (approximately 3 SD from mean). We included participants who had data for both PA score and MVPA PAQ .
Objectively measured physical activity: In a subgroup of participants at schools in proximity to our research centre (n = 9 schools), we measured PA objectively using ActiGraph GT1M accelerometers (Pensacola, FL, USA) with 15-s epochs. Accelerometers were attached to an elastic belt and positioned at the iliac crest. Children were asked to wear the accelerometer during waking hours for five consecutive days (includes weekdays and weekends), only removing it for water-based activities (e.g., swimming, showering). We excluded data from the day of distribution to eliminate between-school differences in distribution time and any initial reactivity to wearing the monitor.
Prior to processing, each accelerometer file was individually screened for spurious data points and patterns. We used custom software (KineSoft Version 3.3.76, Loughborough, UK) and recommended cutpoints [45,46] to classify activity intensity and included all participants with at least 10 h of data on three or more days (no restrictions on weekday vs. weekend days). We considered periods of continuous zeroes ≥ 60 min biologically implausible and excluded them from our analysis. For analysis we used counts/min as an estimate of total PA, and MVPA (MVPA Accel ), defined using a cutpoint of ≥2296 counts/min [45,46]. We also dichotomized children into those that achieved/did not achieve an average of ≥60 min/day of MVPA to estimate compliance to recommended PA guidelines [47].
Cardiovascular fitness: To assess CRF we used a multistage 20 m shuttle run [48] administered in small groups. We recorded the total number of laps each child completed and calculated age-and sex-specific z-scores [49]-we use both the number of laps completed and the z-scores in the analyses. Five schools (two INT, three UP) in northern BC did not complete CRF measurement due to resource constraints.
Anthropometry and demographics: We assessed participants' height without shoes twice to the nearest millimeter using a portable stadiometer (Seca Model 214, Hanover, MD) and body mass twice to the nearest 0.1 kg using an electronic scale (Seca Model 840, Hanover, MD). If measures of height or weight differed by more than 0.4 cm or 0.2 kg, respectively, we obtained a third measure. We used the mean of the two closest values or the median of three equidistant values in our analyses. Body mass index (BMI) was calculated as weight divided by height squared (kg/m 2 ). We estimated maturity offset (years from age at peak height velocity, APHV) using validated equations that include measures of age and height [50]. We determined ethnicity based on parental report of parent and/or grandparents birth place. Most children were white (57%) while remaining children were Asian (25%), North American Aboriginal (10%), and other or mixed ethnicity (9%); this approximates the diversity of the BC population [51].

Sample Size Calculation
AS! BC was powered at the level of teacher implementation (PA delivery) to detect a medium effect size [52] of 0.25 between groups. With a power of 0.80 and α = 0.05 we required a minimum of 64 teachers per condition (128 total; based on a minimum of 3 teachers per school representing 22 schools per condition).

Statistical Analysis
Objective 1-Implementation strategies: We describe the number of implementation strategies and where relevant, the number of activities/events within each implementation strategy.

Objective 2-Implementation measures:
At the teacher level, we assessed differences in PA logs completed by study year (year 1 vs. year 2) and group (INT vs. UP) using Wilcoxon rank-sum tests. We fit linear regression models and adjusted for school cluster (Stata command: vce (cluster)) to compare average PA delivery (min/week) for each year between INT and UP groups. At INT schools, we also compared average Classroom Action PA delivery (min/week) between years 1 and 2.
Objective 2-Impact measures: At the student level, we compared baseline characteristics of students attending UP and INT schools (separately for boys and girls) using unpaired t-tests, ANOVA or Chi-square as required. A month long, provincial teachers' strike (with subsequent student dismissal from school) prevented data collection at 15 schools (9 INT, 6 UP) at the end of year 1. We compared baseline descriptive characteristics of students attending schools that were (n = 15 schools) and were not (n = 15 schools) measured at the end of year 1 using unpaired t-tests.
We compared student-level outcomes between INT and UP schools at the end of years 1 (T2) and 2 (T4) of the randomized controlled trial. For all PA measures and CRF (number of laps completed) we fit sex-specific linear regression models, adjusted for baseline (T1) value, age, BMI, ethnicity and school cluster (Stata command: vce (cluster)). We also replaced baseline BMI with maturity offset [50] in all models to account for potential differences in maturity between groups. The model for CRF age and sex specific z-scores included both girls and boys and did not include age as a covariate but was otherwise similar. For MVPA accel , we included wear time, but it did not improve model fit and was removed from the final model. We assessed model fit visually using model residuals (normality, linearity and homoscedascity) and identified influential data points using Cook's D statistics. We report the adjusted mean difference with 95% CI (INT-UP) and effect size of the regression coefficient for the 'group' variable (Cohen's d; stata command: esizereg [53]) as an indicator of the magnitude of the between-group difference [54]. We describe effect sizes of 0.2, 0.5 and 0.8 as small, medium and large, respectively [52]. For descriptive purposes, we also report unadjusted percent change where appropriate. Finally, we examined between-group differences in the percentage of girls and boys achieving PA guidelines at the end of year 1 and 2 using Chi-square tests; we use Cramer's V as an estimate of the magnitude of the difference and apply the same thresholds as for Cohen's d above. We analyzed student data as per each school's initial random assignment (intention to treat; ITT). However, the largest school in the study (UP; n = 163 students at baseline) spontaneously adopted the AS! BC intervention at the start of year 2. Thus, for the year 2 (T4) analysis we also conducted a sensitivity analysis excluding this school.
Objective 3-Link between implementation and outcomes: We investigated the association between PA delivered (dose) and student outcomes in two ways. First, we used Pearson correlations to describe the association between PA delivered (Total or Classroom Action) and student outcomes at INT schools (no adjustment for school cluster). Second, we dichotomized delivery of Classroom Action PA (the only prescriptive component of AS! BC) into 'high' (≥45 min/week; ≥60% of target [23]) and 'low' (<45 min/week). We then compared change in student-level outcomes between the two groups using linear regression models, adjusted for school cluster as above. All statistical analyses were performed using Stata version 13 (StataCorp, College Station, TX, USA).

Objective 1: Implementation Strategies across 11 Years of Province-Wide Scale up (2004-2015)
We used a variety of strategies across three broad categories (implementation process, capacity-building, scale-up [40]) to support implementation and province wide scale-up of AS! BC across 11 years. We specify implementation strategies in Table 2 and provide relevant data below.  Capacity-building Strategies: involve the management and implementation of a process to deliver AS! BC workshops, build capacity across the province, and maintain interest and momentum within registered schools.

AS! BC workshops and ongoing support
Training workshops were coordinated by the Support Team and delivered by the Support Team, Master or Regional Trainers. Ongoing support (via email, phone) was available as needed.
Actor: AS! BC Support Team Action: The Support Team worked with schools to secure a date for workshops. Classroom Action Training workshops were 3 h long. Refresher training workshops were 1 h long. Training was delivered face-to-face to groups of teachers in schools (gymnasia, multi-purpose rooms, empty classrooms) and followed principles of experiential learning and self-efficacy theory. Training workshops included positive modelling, verbal persuasion and opportunities for teachers to be successful and recognized. The Support Team and Trainers provided ongoing support during the school year via email and telephone, and over time teachers/schools could register for a selection of additional in-person workshops throughout the year (e.g., student playground leadership). Workshop participants completed evaluation forms, which were returned to the Support Team and used to improve future workshops. Target: Participating teachers and schools Temporality: Ongoing throughout the year Conduct ongoing training Make training dynamic Obtain and use feedback AS! BC Master and Regional Trainers The Support Team used a network of Master Trainers and Regional Trainers to deliver workshops across the province. Regional trainers self-identified as interested and submitted a resume. If accepted, they participated in training during a Summer Institute, offered annually.
Actor: AS! BC Support Team Action: The network of Master and Regional Trainers supported the delivery of workshops province-wide.  Regional Trainer Support Materials A number of resources supported the Master and Regional trainers. Key resources included: Regional Trainer Guides (workshop templates, presentation tips, checklists, evaluation forms and handouts for schools); Regional Trainer bags (teaching resources and equipment for leading workshops); AS! BC branded materials (e.g., clothing, clipboard, water bottle, pens); E-news Actor: AS! BC Support Team Action: The Support Team provided Master and Regional Trainers with materials to support workshop delivery around the province Target: AS! BC Master and Regional Trainers Temporality: Master and Regional Trainers received support materials prior to delivering workshops. E-news was delivered monthly.
Provide equipment

AS! BC website
The AS! BC website was updated regularly with registration numbers and contained resources for teachers and schools including: playground circuits, and the latest versions of all support materials and teaching resources.
Actor: AS! BC Support Team Action: The Support Team kept the website up to date with the latest registration data (by school district and school) and resources. Target: Teachers, schools, communities Temporality: Updates and additions to website were ongoing

Develop and distribute educational materials
Scale-up Strategies: The AS! BC Support Team used a comprehensive marketing and promotion strategy to achieve full school district representation and strategically inform community stakeholders to increase support and build sustainable, far-reaching networks for the initiative.

Build partnerships
To build partnerships with schools and stakeholders the Support Team participated in a number of events. This was an opportunity to promote AS! BC, network with relevant stakeholders. Stakeholders spanned the following sectors: government (e.g., ministries of health, education), education (e.g., teachers' associations, parent advisory council, universities), health (e.g., health authorities, regional health units), sport (e.g., sport associations, parks and recreation) and community (e.g., neighbourhood houses, youth organizations).

. Implementation Process Strategies
The Support Team helped schools complete the "4 Steps to Becoming an Action School": register, take stock (needs assessment), take action (action planning), and report. The Support Team set the following targets for initial provincial scale-up in  (2015), >1400 schools (>90%) across BC were registered (involving >87,500 teachers and administrators and reaching approximately 500,000 students).

Capacity-Building Strategies
Capacity-building activities included workshops delivered by the Support Team to train new teachers and maintain interest and momentum within registered schools (e.g., refresher training or workshops on new topics). During scale-up, the Support Team

Scale-up Strategies
Scale-up strategies aimed to increase the number of schools delivering AS! BC. The Support Team used a comprehensive marketing and promotion strategy to achieve full school district representation and strategically inform community stakeholders to increase support and build sustainable, far-reaching networks. The Support Team made presentations to principals and administrators in 40% of school districts during the first year of scale-up; presentations were made to the remaining 60% of districts in the second year. The Support Team also coordinated and/or delivered presentations, displays, and/or promotional materials at events supported by n = 51 organizations in year two of scaleup and n = 64 organizations in year three. Organizations spanned the following sectors: government (e.g., ministries of health, education), education (e.g., teachers' associations, parent advisory council, universities), health (e.g., health authorities, regional health units), sport (e.g., sport associations, parks and recreation) and community (e.g., neighbourhood houses, youth organizations).  (Figure 2). Due to teacher and student movement between classes after year 1, 128 teachers (72%) participated during 1 year only (59 in year 1 only, 69 in year 2 only); 51 teachers (28%) participated across both years. Therefore, 110 teachers participated in year 1, and 120 teachers participated in year 2.

Randomized Control Effectiveness
We provide baseline student characteristics in Table 3. Age, BMI, maturity offset, PA score and CRF were similar between INT and UP girls at baseline. Girls attending INT schools had higher MVPA (self-reported and accelerometer-measured) and total PA and were more likely to achieve PA guidelines than UP girls. Boys attending INT schools were younger, further from APHV (less mature), and less fit than UP boys. However, MVPA PAQ was lower among UP boys. BMI, PA score, MVPA Accel and the proportion meeting PA guidelines was similar between INT and UP boys. In both girls and boys there was a higher proportion of white vs. Asian students at INT schools compared with UP schools.
Baseline BMI, PA score and CRF were similar between girls who were measured at the end of year 1 and those who were not (due to the provincial teachers' strike). However, girls not measured at the end of year 1 were younger (−0.1 year, p = 0.007) and reported higher MVPA PAQ (+20.8 min/day, p = 0.009) at baseline than girls that were measured. Age, BMI, and CRF were similar between boys who were measured at the end of year 1 and those who were not. However, boys not measured at the end of year 1 had higher PA scores (+0.1 units, p = 0.03) and MVPA PAQ (+32.2 min/day, p = 0.001) at baseline than boys at measured schools. Finally, there were more Asian students, and fewer white students at measured schools compared with non-measured schools (% Asian/white/other: 38/45/17 vs. 7/75/18, p < 0.001). This was a function of focusing our limited T2 measurement (due to teachers' strike) on schools in proximity to our research centre in Metro Vancouver.

Physical Activity Delivery by Teachers
Twenty-three teachers in year 1 and 11 teachers in year 2 did not return any activity logs. A further 20 teachers (1 in year 1 and 19 in year 2) completed their logs retrospectively and 11 teachers (all year 2) completed a mix of on-time and retrospective activity logs that we could not disentangle. On average, teachers with one or more retrospective logs reported delivering more weekly PA than those completing all logs on time (+17.1 min/week; 95% CI 4.2, 30.0). As a result, we include only those with exclusively 'on time' logs in these analysis (n = 86 in year 1 and n = 79 in year 2). These teachers returned a median of 81% activity logs (IQR 57-90%) in year 1 and 77% activity logs (IQR 54-94%) in year 2. Teachers at INT schools returned more logs than did teachers at UP schools (median 86% vs. 70% for INT and UP, respectively). Table 4 summarizes total PA delivery by year and group assignment for teachers with at least one 'on time' PA log (year 1: 78%, year 2: 66%). In year 1, INT teachers delivered more PA as compared with UP schools. In year 2, PA delivery was similar between INT and UP schools. Within the ITT analysis, PA delivery did not differ by year within UP or INT schools. However, when we excluded 5 influential data points (PA delivery well above (178-278 min/week) or below (22.5 min/week) the average (122.5 min/wk)) PA delivery at INT schools decreased from year 1 to year 2 (year 1 vs. year 2: −16.7 min/day; 95% CI −32.0, −1.4). Teachers at intervention schools who participated across both years of the study provided +25.4 min/week more PA during year 2 than teachers who were new in year 2. Overall findings did not change with the sensitivity analysis (data not shown). Within INT schools, teachers delivered a similar amount of Classroom Action PA between years 1 (33.1 min/week; 95% CI 25.3, 41.0) and 2 (25.9 min/week; 95% CI 13.7, 38.1).    UP, Usual practice; INT, intervention; BMI, body mass index; APHV, age at peak height velocity; PA, physical activity; MVPA PAQ , Moderate-to-vigorous PA from question 1 of the Physical Activity Questionnaire-Child; MVPA Accel , Moderate-to-vigorous PA from accelerometer. Difference calculated as UP-INT; values in bold indicates significant difference between UP and INT within sex; † Significantly different from girls. Note: We excluded 13 students (4 INT girls, 5 INT boys; 1 UP girl, 3 UP boys) from analysis who had medical conditions that interfered with participation in regular PA or cardiovascular health (e.g., cerebral palsy, diabetes, juvenile arthritis, cardiac anomalies, spina bifida). We also excluded 175 participants missing data for one or more covariates: age (

Student's Physical Activity and Fitness
Self-reported PA (PAQ-C): PA score was similar (effect sizes <0.1), between INT and UP girls at the end of years 1 and 2 ( Table 5). PA score was lower among INT boys as compared with UP boys at the end of year 1, but not year 2 (Table 5); however, effect sizes were small (~0.2). Sensitivity analysis did not change girls' results; however, boys at INT schools had lower PA score at the end of year 2 than boys at UP schools within the sensitivity analysis (−0.2; 95% CI −0.3, −0.02; (Supplementary Table S1)). For girls and boys, MVPA PAQ did not differ between INT and UP groups at the end of year 1 or 2 within ITT or sensitivity analyses (effect sizes all <0.2). We provide the intraclass correlation coefficient (ICC) for baseline, change from T1-T2, and change from T1-T4 in Supplementary Table S2. Replacing baseline BMI with maturity offset at baseline did not change results.
Objectively measured PA (accelerometer): For girls and boys, total PA and MVPA Accel were similar (effect sizes range from −0.16 to 0.05) between INT and UP groups at the end of years 1 and 2 (Table 5). Sensitivity analysis did not change boys' results; however, girls at INT schools had lower total PA and MVPA Accel at the end of year 2 compared with UP girls (Supplementary Table S1). More INT girls than UP girls met PA guidelines at the end of year 1 (40% vs. 22%, p < 0.001, effect size = 0.19); however, achievement of PA guidelines was similar at the end of year 2 (33% vs. 22%, p = 0.06, effect size = 0.13). Achievement of PA guidelines did not differ between INT and UP boys at the end of year 1 (48% vs. 55%, p = 0.3) or year 2 (48% vs. 51%, p = 0.7) (effect sizes < 0.1 for both). We provide the intraclass correlation coefficient (ICC) for baseline, change from T1-T2, and change from T1-T4 in Supplementary Table S2. Replacing baseline BMI with maturity offset at baseline did not change results.
Cardiorespiratory fitness: On average, girls at INT schools completed 7.5 more laps (95% CI −0.5, 15.5) than girls at UP schools at the end of year 1 (effect size = 0.48; Table 5). The increase in number of laps corresponded to a 26% increase in girls at UP schools, and a 62% increase in girls at INT schools (unadjusted percent change). On average, boys at INT schools completed 4.7 more laps (95% CI 1.4, 8.0) than boys at UP schools at the end of year 1 (effect size = 0.28; Table 5). The increase in number of laps corresponded to a 23% increase in boys at UP schools, and a 50% increase in boys at INT schools (unadjusted percent change). The between-group difference was <1 lap (effect sizes < 0.1) at the end of year 2 for both girls and boys (Table 5). On average, age-and sex-specific z-scores were 0.33 greater (95% CI 0.2, 0.5) in children at INT schools compared with UP children at the end of year 1 (effect size = 0.36); there was no between-group difference at the end of year 2 (−0.03; 95% CI −0.3, 0.2; effect size < 0.1). Sensitivity analysis did not change the findings (Supplementary Table S1). We provide the intraclass correlation coefficient (ICC) for baseline, change from T1-T2, and change from T1-T4 in Supplementary Table S2. Replacing baseline BMI with maturity offset at baseline did not change results. Table 5. Physical activity (PA) and cardiorespiratory fitness (CRF) at the end of year 1 (T2) and 2 (T4) in girls and boys attending usual practice (UP) compared with intervention (INT) schools. Values are means (SD) or adjusted mean difference (95% CI) using an intention to treat analysis. We include effect size (ES; Cohen's d) as an indicator of the magnitude of the between-group difference.

Link with Implementation
Dose of Total PA delivered was positively associated with change in MVPA PAQ during year 1 (r = 0.17, p = 0.02). Dose (Total or Classroom Action PA) did not correlate with change in any other student-level outcome (Supplementary Table S3). Change in student outcomes did not differ between children in classes that received at least 60% delivery of Classroom Action PA compared with classes exposed to <60% in years 1 or 2 (Supplementary Table S4).

Discussion
Researchers advocate for whole-of-school [55] and scaled up approaches [19], yet a dearth of effective whole school PA interventions were scaled-up [24]. There are myriad reasons why scale-up does not occur. Not least of which is the considerable support, knowledge and sustained resources needed to do so. Thus, we felt it important to chronicle the phased approach we adopted to implement and scale-up Action Schools! BC across 11 years, with committed support from BC government ministries and a host of school community partners (Figure 1). We also describe implementation and scale-up strategies that facilitated implementation and sustained AS! BC delivery, as this may aid other teams working in this important field of study.
Schools are complex and dynamic systems [56], and present many challenges to scaling-up effective interventions. The rapid uptake of AS! BC in the first three years of scale-up illustrated the need and appetite for the intervention, a window of opportunity [57] afforded by the political climate and the success of our scale-up and implementation strategies. It was not possible to formally evaluate the influence of individual strategies. However, we note that our implementation and scale-up approach incorporated many 'essential elements' of comprehensive school health implementation models [58]. Our approach also aligned with implementation and scale-up frameworks [23,36,37], and what policy-makers deemed the drivers of successful scale-up and sustainability [57]. Governance, leadership, resources, outsourcing delivery, accountability structures and committed stakeholder engagement all played a key role to support scale-up and sustainability of AS! BC.
Much has been written about the key role of central support teams to build capacity and provide ongoing support to increase deliverers awareness, knowledge, skills, selfefficacy, and motivation to adopt and implement effective interventions [36,59]. Capacitybuilding studies in public health and community-based practice identified technical assistance, training and tools as central strategies to support effective implementation [23,36,60]. These were all focal points of AS! BC, and received sustained investment from BC government ministries. For example, grants to support scale-up funded the Support Team (5-8 full and part-time staff and various contractors; their office space; travel; communications, etc.), covered costs for teachers and facilitators who delivered AS! BC workshops (travel, training, release time or honoraria as needed), AS! BC resources such as equipment (Classroom Action) bins, resource manuals and other printed materials required for day-to-day delivery of the model, and release time for classroom teachers to attend training workshops.
Unique features of AS! BC also likely played a role in successful scale-up. The multicomponent model provided schools flexibility to create their own Action Plans, based on teachers' and schools' available resources and self-identified needs. This provided schools autonomy to adapt the model to context-a core condition for successful implementation of a comprehensive school health approach [58]. New implementation strategies and program materials were added over time to address the changing needs of schools, and sustain interest. For example, after the effectiveness trial, the Support Team added new workshops on healthy eating and PA across all grades, arranged teacher mentorship, provided refresher and student leadership training and offered additional resources such as e-newsletters, posters and instructional manuals. The Support Team tracked delivery of each of these; we did not assess outcomes related to these additional activities.
Below, we delve further into the unique aspects of AS! BC scale-up, and present key strengths, limitations and implications of our findings.
Findings from the subset of schools that participated in the two-year randomized, controlled effectiveness trial showed that teachers at intervention schools delivered more PA during year 1 of the study as compared with teachers at usual practice schools. However, teacher compliance with activity logs was a challenge. Others noted this previously and attributed poor compliance to the substantial demands on teachers across a broad range of administrative and instructional tasks [61]. Despite greater PA delivery by teachers in year 1, students' PA did not increase significantly. This is consistent with some [22,62], but not all [63,64], systematic reviews published in the last 10 years. Even when improvements in students' PA were observed in other trials, the effect was modest. For example, across 20 controlled trials students' PA (moderate and vigorous) increased by approximately one min/day [63].
Nevertheless, in the current study, for intervention schools we observed a 36% and 27% (unadjusted percent change) greater increase in girls' and boys' fitness, respectively, compared with peers attending usual practice schools during year one. While this was not statistically significant after controlling for the relatively small number of variably sized clusters, we recognize the limitations of relying solely on p values [54]. This between-group difference corresponds to approximately 30-60 s additional running or a difference in peak oxygen consumption of approximately 1.0 mL/kg/min for boys and 1.9 mL/kg/min for girls at the end of year one-two to four times greater than that observed in a recent pooled analysis of 20 controlled trials [63]. We consider the magnitude of change indicated by the meaningful effect sizes, particularly in girls, an important finding. On average, girls are less fit compared with boys-a difference which increases with age and maturation [65]-and, tend to benefit less from interventions than boys [63]. We also consider this result promising in light of the suggested secular decline in children's fitness of 7.3% between 1981 and 2014 [7]. While long-term implications and clinical relevance of small improvements in children's CRF are difficult to discern, evidence in adults suggests that even small improvements in CRF yield clinically important outcomes [66]. For example, in adult men, a 1-minute increase in treadmill test duration over approximately 5 years was associated with a 7.9% reduction in all-cause mortality and an 8.6% reduction in cardiovascular disease mortality [66]. As CRF tracks through childhood [10] and adolescence [11] there is a potential for these small changes to influence adult health.
Our previous efficacy trial improved PA [29] and CRF [30] among students who attended AS! BC schools. This pattern, wherein benefits decrease with the move from efficacy to real world trials, is not unique to our work [67]. Scale-up requires adaptation to fit new contexts and delivery systems, and to accommodate resource constraints. This may lead to an attenuation of the intervention effect-known as the "scale-up penalty" [68,69] or an intervention "voltage drop" [70] of 25-50% [67]. In the effectiveness trial, intervention teachers delivered~14-33 min/week more PA than teachers at usual practice schools; the difference in PA dose between teachers at intervention compared with usual practice schools was higher (45-55 min/week) in our efficacy trial [34]. In the Classroom Action zone, teachers in the effectiveness trial delivered~47% of the target PA in Year 1, and~35% of the target in Year 2. An estimated 60% of program delivery is required to elicit beneficial change [23]. Therefore, at AS! BC scale-up, dose of PA delivered may have been insufficient to elicit similar improvements in children's PA and fitness, as compared with the efficacy trial. Our findings highlight the challenge of implementing an 'effective' intervention to accommodate scale-up. This 'tug-of war' [71] to retain implementation fidelity while adapting school-based interventions to achieve 'best fit' for diverse geographic regions and school populations, is a topic that deserves further attention.
A number of factors may have contributed to voltage drop at scale-up in our study. First, due to a provincial teacher strike, we were unable to collect data in some schools at the end of year 1. This reduced our sample size for key outcomes and impacted our analyses. Second, we decreased the amount of support provided to teachers during the effectiveness trial (as compared with the efficacy trial), as the Support Team sought to reduce intervention delivery costs during scale-up. Third, intervention teachers delivered less PA in year 2 compared with year 1. This might reflect that training sessions in year 2 were 'refresher' workshops (~1 h duration) compared with full workshops in year 1 (~3 h duration). Thus, new teachers in year 2 may not have been as well-trained or committed to deliver the AS! BC intervention as teachers in year 1. Teachers at intervention schools who participated across both years of the study provided +25.4 min/week more PA during year 2 than teachers who were new in year 2. We [20] and others [23] have shown a positive relationship between dose and at least one health outcome. We noted a positive association between total PA delivered by teachers and change in students' PA score over the first year of the study. This highlights that engaged, motivated, and trained teachers are key to successful outcomes at the student level. Thus, as a means to continually build capacity, teacher training through ongoing (new and refresher) courses each year is a worthwhile investment. Fourth, unlike the efficacy trial, there were no self-identified champions at schools during scale-up. Scale-up models [72] identify the key role that local champions play in scale-up success.

Strengths and Limitations
We highlight several strengths of our study that address gaps in the current literature. Specifically, (i) we describe the chronology of AS! BC scale-up, and implementation strategies that supported its translation into a "real-world" setting. Across school-based health promoting interventions more broadly, only 5 examined sustainability more than 5 years later [25]. To our knowledge, AS! BC is one of only two other whole-school PA models (CATCH [73] and Take10! [74]) to be implemented at scale with continuous stakeholder support over more than a decade; (ii) we evaluated the impact of AS! BC across two years of scale-up in a large and diverse cohort of children (~1500 children across the province); and (iii) we investigated the link between program implementation and student-level health outcomes.
There are many challenges associated with conducting pragmatic, real-world trials in schools. Therefore, we acknowledge several limitations of our study. First, due to variability in the number of children per school (cluster), and a teacher strike that rendered us unable to measure 50% of participating schools at the end of year 1, we could not apply multilevel modeling techniques to account for our clustered study design [75]. However, we adjusted the standard errors to account for the non-independence of individuals from the same school cluster [76]. This approach may have limited our ability to detect statistically significant intervention effects, despite large changes in CRF (Type 2 error). Second, due to poor compliance with activity logs we could not determine the extent to which teachers participated in all aspects of intervention delivery (i.e., across all Action Zones). Third, without direct observation of classroom teachers, and subsequent student participation (not feasible in scale-up studies), we do not know the intensity of PA delivered (dose delivered), or the degree to which students responded to, and were engaged in, the activities (dose received). We recognize the distinction between 'dose delivered' by teachers and 'dose received' by students and aspects of implementation that influence these implementation indicators (e.g., fidelity, quality and participant responsiveness) [23], that we did not measure.

Conclusions
It is possible to scale-up and sustain whole of school PA interventions over the longer term (>10 years), with the ongoing support of government and school-community stakeholders. Despite many challenges to doing so, there is a need for implementation resource teams to conceive of new ways to sustain benefits of scaled-up school-based health promoting interventions to improve students' health at a population level. Support units with established connections to schools and comprised of researchers, government stakeholders and school community-based practitioners are a critical element of scale-up success. Greater PA delivery by trained teachers and the magnitude of change in fitness we observed in AS! BC intervention schools, suggests that scaled-up school-based PA models may enhance children's health. Notably, implementation and capacity-building through ongoing teacher training, support and compliance are key factors to realize and sustain benefits. Future school-based interventions guided by rigorous implementation process models that support program delivery, and evaluation frameworks that assess key implementation, scale-up and sustainability indicators would be a welcome addition to the current literature.  Informed Consent Statement: Informed consent was obtained from all participants (parents/guardians as needed) involved in the study.

Data Availability Statement:
The datasets used during the current study are not publicly available as stipulated in our participant consent forms but are available from the authors on reasonable request.