The evaluation of subject teachers’ experiences and perceptions of physically active academic lessons in general upper secondary school was carried out in the context of the development project Moving Academic Lessons in General Upper Secondary School (2022‐2024). The aim of the development project was to promote the learning, well-being, and health of general upper secondary school students by combining PA with academic lessons. This development project was funded by the Ministry of Education and Culture and implemented for general upper secondary school teachers in Finland in cooperation with a national action program, Finnish Schools on the Move.

Fourteen subject teachers (n=14) from 8 different schools accepted an invitation to participate in the development project as PAL pioneer teachers in spring 2022. Participants represented 16 different teaching subjects, and their teaching experience in general upper secondary education ranged from 2.5 to 29 years, with a mean of 12 years. The subject teachers were asked to use and create a variety of teaching methods to suit PA with their own subject lessons’ topics and contents, and implement PA breaks during academic lessons. The development project was reviewed to inform the formulation of teacher interview questions (Part 1 of the PAAL study) and the planning of teaching methods for the intervention (Part 2 of the PAAL study).

Fourteen PAL pioneer teachers involved in the development project were invited to participate in the PAAL study teacher interviews conducted between October 2023 and May 2024, and all invited teachers (13 females and 1 male; ages 30‐62 y) accepted the invitation. Participants received correct and sufficient information about the study, including its objectives and the intended use of the interview data.

Participation was voluntary, and participants were informed of their right to withdraw from the study at any time. Written informed consent was obtained from all participants prior to the interviews. Privacy and confidentiality were assured, and no financial or other compensation was offered. All identifying details have been omitted from the published written descriptions to protect the anonymity of the participants.

Before data collection (September 2023), the need for an ethical review of this part of the study was requested from the Ethics Committee of Human Sciences of the University of Oulu. According to the statement of the Ethics Committee of Human Sciences of the University of Oulu, an ethical review is not required when research involving human participants complies with the 2019 guidelines of the Finnish National Board on Research Integrity (TENK) [54]. This study was conducted in accordance with TENK’s guidelines. In addition, before teacher data collection, the study was approved by the Education and Culture Services of the municipality where the data were collected (OUKA/11121/07.01.04.02/2023) and fulfilled the ethical standards for empirical research.

The participating subject teachers were interviewed after implementing PAL and PA breaks in their teaching for more than 12 months. The interview questions were finalized after a pilot interview. The list of final interview themes was shared with participants 2 days before the interview. Interviews were conducted at times and locations convenient for the teachers: at the participant’s school, at the university, or at the interviewer’s home. Teachers were asked to describe their experiences with PAL and PA breaks. The questions were organized around the main themes (see Teacher Interviews in Multimedia Appendix 1), which were developed through discussions with PAL pioneer teachers and reflections from the development project. In addition, similar overarching themes have been identified to summarize teachers’ perspectives on implementing PAL in primary schools [55]. Although the interviews adhered to the main themes, the order of questions varied between interviews based on the flow of the conversation between the interviewer and the interviewee. The semistructured interviews were recorded and transcribed during the autumn of 2023 and the spring of 2024 in Northern Finland. Interviews lasted 63 to 130 minutes (mean 95 min) and were done individually. Together, there were 365 pages of long transcribed text. PAL and PA breaks were treated as distinct classroom activity types. The interview data were analyzed using thematic analysis [56]. An inductive approach was conducted through 6 phases [56] (see Analysis of Teacher Interviews in Multimedia Appendix 1) to gain insight into why teachers integrate PAL and PA breaks into academic lessons. In addition to the inductive approach, a deductive approach was used to identify teachers’ perceptions related to successful PAL adoption and implementation.

The effectiveness of physically active academic lessons among students was examined using a cluster randomized individual crossover trial conducted in two waves. In spring 2024, a total of 4 teachers of mathematics and foreign languages from 5 general upper secondary schools in Northern Ostrobothnia, who had participated in both the development project and the teacher interviews, were invited and agreed to implement the intervention. In autumn 2024, 2 schools from Central Finland were contacted, and a total of 5 teachers of mathematics and foreign languages from these schools expressed interest and agreed to implement the intervention. From a total of 17 eligible course groups of first- and second-year students, 168 students volunteered to participate in the study. These groups included 10 from schools in Northern Ostrobothnia (4 in mathematics and 6 in foreign languages), comprising 242 students, of whom 74 volunteered, and 7 from schools in Central Finland (4 in mathematics and 3 in foreign languages), comprising 200 students, of whom 94 volunteered. Students had the opportunity to withdraw from the study at any stage, and any data collected up to the point of withdrawal were retained as part of the research material. Eight students dropped out of the study.

Students in the same course group formed the clusters (mean cluster size 10, ranging from 1 to 18), which were randomized to undergo the 3 different treatments (lessons) in a different order. As there are 6 possible permutations of the 3 treatments, each cluster was allocated one of these sequences using a random number generator that sampled integers uniformly from 1 to 6. This procedure ensured randomized assignment of treatment order across the groups. After the trial, 30 voluntary students were invited to student interviews. This study followed the Consolidated Standards of Reporting Trials (CONSORT) statement (2025). The CONSORT flow diagram of the trial is presented in Figure 2.

The study protocol was approved by the Ethics committee of Jamk University of Applied Sciences (Ethical review statement 472673, 21 December 2023). The study was prospectively registered (ISRCTN63809854, registered 11 January 2024 [57]).

Informed consent was obtained from the students and, in the case of minors, also from their guardians. This included consent for data collection as described in the information sheet and for the use and processing of personal data in accordance with the privacy notice. The consent form further included consent to being contacted after the trial for the purpose of inviting the students to a potential interview. Participation in the study was voluntary, and no financial or other compensation was offered for participation in the trial. Students who voluntarily chose to participate in the post‑trial interview were offered a €10 gift card (approximately US $11) as compensation.

With regard to privacy and confidentiality, all study data were pseudonymized prior to analysis, with personal identifiers replaced by ID numbers and accessible only to data managers/statisticians. All data are stored on secure, access‑controlled local servers at Jamk University of Applied Sciences with daily backups. Access is restricted to authorized research team members and managed by the principal investigator. No identifiable information will be reported.

The first wave of the trial was delivered in spring 2024, and the second wave of the trial was delivered in autumn 2024. The trial focused on 2 subjects: mathematics and foreign languages, representing different types of subjects. Students participated in 3 different treatments (lessons) in random order. All lessons followed the National Core Curriculum for General Upper Secondary Education, but the teaching methods were different:

A description of the PAL methods and PA breaks used for treatment 1 and 2 is described in more detail (Textbox 1). The content of these methods was constructed in collaboration with general upper secondary school teachers, teacher educators, PA and learning researchers, and students.

Researchers held a 2-hour meeting to present the study, PAL method principles, and PA breaks to the teachers implementing the intervention. The PAL method (treatment 1) was similar for every group, but the lesson-specific content was designed by the teacher. This ensured that the lesson content was a part of the course learning objectives. Teachers had the opportunity to ask for assistance from research team members and were required to submit lesson plans. The PA breaks (treatment 2) were the same for every group, and teachers only needed to play them on the screen for the students. Because only some of the teachers and students were familiar with the different physically active classroom practices, the teachers were instructed to rehearse the PAL method and PA breaks with the students to familiarize themselves and the students with the methods before the measurements.

Given that the physically active lessons adhered to the national curriculum and consisted of tasks typical of students’ everyday lives, participation did not pose any risk beyond that of a regular school day. Despite the study’s hypotheses, PAL and PA breaks may cause negative effects, such as reduced situational engagement, alertness, or executive functions. These effects were measured and analyzed as part of the intervention outcomes, and subgroup analyses may reveal whether modifying factors are associated with perceived disadvantages.

The fidelity of the intervention was monitored by having each lesson observed by trained personnel, who recorded the course of the lesson and any deviations or unusual events. Apart from the addition of the second wave, no changes were made to the study protocol after trial registration, and no major deviations from the protocol occurred during the study.

The majority of the measurements were performed as part of the lessons conducted by trained personnel. In the baseline measurement session, students’ executive functions (inhibition and working memory) and academic skills such as reading and math fluency and spatial visualization were measured. In addition, the devices (Axivity AX3 and Firstbeat Bodyguard 3 monitoring device) and instructions for measuring PA, sedentary time, stress, recovery, and sleep were distributed, and instructions for self-monitoring usual daily and hourly variations in alertness were given. Finally, personalized answer links to complete the background survey were sent to the students via email. Students completed the questionnaire in their free time.

Students wore the devices for two 4-to-5-day measurement periods (including the treatments). The first measurement period started straight after the baseline measurement session, or at the latest, in the morning 3 days before the first intervention lesson, and it ended in the morning after the first intervention lesson. The second measurement period started in the morning of the day before the second intervention lesson and ended in the morning after the last intervention lesson. Students also filled out an online alertness questionnaire 4 times per day (9 AM, 11 AM, 1 PM, and 3 PM) for 3 days. The timeline of the measurements is presented in Table 1.

Measurements during the intervention lessons included a short initial survey, an executive function test, and an alertness questionnaire at the beginning of the lesson; 2 alertness questions during the lesson; and an alertness questionnaire, an executive function test, and a situational engagement questionnaire at the end of the lesson (Figure 3). The initial survey (2 minutes to complete), conducted on students' own computers through Visual Learning Environment (ViLLE), a widely used digital assessment and learning platform, included questions about the quality of the previous night's sleep, as well as the consumption of breakfast, lunch, caffeine, nicotine products, and PA before the treatment. Students also wore the Axivity AX3 and Firstbeat Bodyguard 3 monitors to assess PA, standing, sitting, stress, and recovery during the intervention lessons.

To monitor physiological stress and recovery, students recorded a single-channel electrocardiogram to measure heart rate variability (HRV). HRV is measured as the time interval, in milliseconds, between two consecutive R-wave peaks (RR-Interval) in the electrocardiogram. The measurement was done using a Bodyguard 3 (Firstbeat Technologies Oy, Jyväskylä, Finland) and the Firstbeat Life app. The device was attached with self-adhesive single-use electrodes (Arbo H92SG). Participants wore the device around the clock, except for showering and water exercise. In the app, the students logged their activities in the supporting electronic diary.

Data were processed using Firstbeat research software. Firstbeat Life analysis is based on heart rate analysis software that identifies different physiological changes through HRV. Physiological recovery, stress, sleep, and PA were collected [67]. A score ranging from 0 to 100 was generated for each of the different physiological states for each day (morning to morning). A higher number represents a better score.

Mental load was assessed using questionnaires measuring study burnout and engagement as part of the background survey. The Student Burnout Inventory-9 [68] assesses study burnout through 9 items measuring exhaustion, cynicism, and inadequacy at school, rated on a 6-point scale (1=strongly disagree; 6=strongly agree). Meanwhile, Energy, Dedication, and Absorption-9 [69] assesses study engagement through 9 items, rated on a 7-point scale (0=not at all, 7=daily).

Perceived physical competence was measured as part of the background survey using a questionnaire that included a modified Finnish version of the Perceived Physical Competence Scale for Children [70] and questions about the ability and desire to develop one’s own physical ability [71]. Altogether, they comprise 12 items rated on a 5-point scale.

Perceived academic competence was assessed using a questionnaire including self-reported academic achievement score and self-perceptions (in math and foreign language) [72]. Perceived academic achievement score was assessed with a question “How would you rate your current academic performance in mathematics?” rated on a 4-point scale (1=excellent, 4=poor). Self-efficacy in math and foreign language was measured with 4 items (eg, “Mathematics/foreign language is an easy subject”) with a 5-point Likert scale (1=completely agree, 5=completely disagree).

To measure PA and sedentary time, students wore an accelerometer (AX3 [73]) on the anterior midline of either thigh. The accelerometer was attached to the skin using Opsite Flexifix transparent adhesive film (8 × 10 cm), with a piece of mesh between the monitor and skin to enhance breathability [74]. Envelopes including the monitor, detailed instructions including video link for how to attach the monitor, 5 pieces of adhesive film, and mesh swabs were delivered to participants face-to-face in a classroom. Data were collected with a 100 Hz sampling rate and a dynamic range of ±8 g.

Raw data were downloaded using OMGUI (version 1.0.0.45; Open Lab) [75]. ActiPASS (version 1.62) [76] was used to perform activity classification. Data were processed using the software’s default ProPASS settings, which have been validated to work well with children’s data as well as adults’ [77]. The start and end times, time at school, and intervention class times for each participant’s data were imported to ActiPASS using a diary file to inspect different time domains separately. The autocalibration function of the software was used. Activity was classified based on movement type: sleep, lie, sitting, standing, walking, running, stair-walking, cycling, but also based on the intensity of movement: sedentary, light, moderate, and vigorous PA.

To address baseline cognitive functions and academic skills, performance in the following tasks was evaluated through the ViLLE platform: the Choice Reaction Time task (as a warm-up task), the Sentence Reading Fluency task (assessing reading fluency), the Calculations task (assessing arithmetic fluency), the Digitized Corsi Blocks task (assessing working memory), the Spatial Relations subtask of the Woodcock-Johnson III Cognitive Assessment (assessing spatial visualization ability), and the Flanker task (assessing inhibition). Completing the tests took approximately 20 minutes. The tasks are described in more detail in Table S1 in Multimedia Appendix 1.

The background survey was conducted electronically using Webropol software (Webropol Ltd, Helsinki, Finland), consisted of 55 questions and took a median time of 20 minutes to complete. The collected data included a comprehensive list of background variables: personal identifying information, age and sex, perceived health, height, weight, physical fitness, several questions on PA, sleep and wake-up times, nutrition, family form, parental socioeconomic questions, background academic performance in comprehensive school, and language syllabi in comprehensive school. Regarding the studies in upper secondary school, the questionnaire measured, for example, perceived performance and motivation in studies and diagnosed learning difficulties.

Generalized linear mixed-effects models will be applied to study the acute effects of the intervention on outcome measures. A cluster-randomized individual crossover design is assumed, that is, individuals serve as their own controls when the clusters (school classes) are crossed over the three intervention arms. This is the most efficient of cluster-randomized study designs [78]. The data will be analyzed according to intention-to-treat principles. Data lost due to participation withdrawals, student absences from measurement lessons, and missingness caused by other reasons will be multiply imputed in the analysis of primary outcome variables.

Within the analysis of main outcomes of classroom interventions (alertness, executive function, and situational engagement), the first analyses will quantify the main intervention effects on the outcomes. The main effects in generalized linear mixed-effects models will be adjusted for 3 levels of confounding factors. First, the factors that are consequences of the study design are taken into account (season, lesson time, subject, and group ID as a random intercept). Second, the student-level constant factors (sex, learning-related disabilities) will be included in the models. Finally, the third set of models will be additionally adjusted for individual lesson-specific factors (prelesson activities, baseline alertness, quality and duration of preceding sleep, preceding meals, smoking and caffeine intake, and menstruation status). The further manuscripts will study modifying factors for the intervention effects. The studied modifying factors are planned to include current physical and mental load, perceived physical competence, and perceived academic competence. The models will include corresponding or less extensive adjustments based on the observations of analysis of main intervention effects. Mplus and R statistical packages will be used.

Prerecruitment power calculations were conducted to determine the required sample size for the study. Required statistical power (1-β) was set to 90%, and significance level (α) was set at .05. The study was planned to detect medium intervention main effects (Cohen d=0.50) in the primary outcome variables: situational engagement, alertness, and inhibition. Definition of the desired minimum detectable effect size was, as usual, a compromise between academic desire and study costs and supported by published works on the outcomes (eg, alertness [79,80], inhibition [81,82]).

The required sample size n for an individually randomized controlled trial would be

nIRCT=(Z1−α/2+Z1−β)2⋅2σ2Δ2,

where Zx is the xth quantile of the standard normal distribution, α is the size of the test (ie, level of significance), β is 1 – required level of statistical power, Δ is the minimum detected difference in the levels of outcome, and σ2 is the assumed overall variance of the outcome variable.

In educational interventions, participants are most often divided into school classes, and therefore, individual randomization for intervention groups is rarely feasible. This cluster-randomization weakens the design generally, and the required sample size increases by a factor called the design effect.

In our study design, individuals serve as their own controls by visiting the two intervention treatments and a control treatment in randomized order. The design effect gets a form of

DECRIC=2(1−ρ1−ρ2),

where ρ1 is the assumed intra-subject correlation and ρ2 is the assumed intra-cluster correlation.

Assuming the variance structure as observed in the pilot study (Matikkavire), ρ1=0.31 and ρ2=0.01, the design effect is 21-0.31-0.01=1.36 and the required sample is n=115 individuals to be crossed over the treatments. By assuming 33 students per class and a 70% student participation rate, we need to recruit five classes. To balance the design, we decided to recruit six classes such that two classes would have been randomized to each of the three intervention arms, where the order of treatments is rotated. As the number of volunteering participants per group remained lower than expected, more groups were invited, and eventually, the recruitment continued in a second wave in the autumn of 2024 to obtain the desired statistical power.

The forthcoming subgroup analyses (eg, sex-specific analyses) and moderation analyses (moderating factors will include physiological and psychological state, and physical and academic competence of the student) have less statistical power to detect medium interaction effects. Statistical power is approximated for subgroups by standard methods as discussed above, and for more complex moderation analyses by, eg, post data-collection Monte Carlo simulation. The post hoc power calculations are not used to determine sample size, but to support the confidence of the interpretation of the results of the moderation models.