Randomized controlled trial aimed at analyzing the effectiveness of a 12-week yoga-based intervention on perceived stress, emotional wellbeing, state anxiety and trait anxiety in university students. This study was registered prior to its commencement (NCT06304662) and was approved by the Ethics Committee of the Mid-Atlantic University (CEI/05–006).

The recruitment of participants was carried out from November to February of 2024, taking advantage of placing advertisements at key points within the university campus, dissemination through social networks linked to the university, and direct contact with students via emails sent from the educational entity’s database. Those who showed interest in participating in the study completed an informed consent form in accordance with the Declaration of Helsinki, Good Clinical Practice, and all applicable laws and regulations, which clearly explained the study’s goals, the methods to be followed, potential risks and benefits, and the confidential handling of the obtained data. Included in the study were students who: (i) were university students who had participated in a yoga training program in the previous 12 months; (ii) have adequate physical capacity to carry out the activities demanded in the study; and (iii) that they have the ability to interpret the project instructions, programs and protocols. The exclusion criteria were: (i) that they did not meet the minimum number of sessions required to complete the study; and (ii) that they had extensive experience and experience in the practice of yoga.

The distribution of participants was carried out using a sequence of random numbers generated by computer, placing individuals equally in an experimental group (EG) and a Control group (CG) following a 1:1 ratio. The group assignments were kept hidden from both the participants and the researchers, and the physiotherapist responsible for carrying out the intervention. This process was carried out using sealed, opaque envelopes that were sequentially numbered, which were kept locked and were only opened by a person not directly involved in the study. At the conclusion of this procedure, 65 individuals were assigned to the experimental group and another 65 to the Control group.

Participants in the experimental group (EG) underwent a yoga program twice a week (Tuesdays and Thursdays) for 12 weeks in person, totaling 24 sessions with duration of 60 min each. Each of these sessions during this period consisted of a total of 4 distinct parts: (i) 5 min dedicated to breathing techniques through Pranayama postures; (ii) 10 min of warm-up based on joint movement exercises; (iii) 35 min allocated to the central part of the intervention where postures based on Hatha yoga were performed, one of the most traditional and widely practiced forms of yoga that focuses on physical postures (asanas), breathing techniques (pranayama) and meditation (dhyana); and (iv) 10 min for relaxation techniques based on flexibility and stretching exercises. In the core part of the intervention, different postures were carried out distributed according to position: (a) standing postures: Utkatasana (Chair Pose), Uttanasana (Standing Forward Bend), Vrksasana (Tree Pose), Garudasana (Eagle Pose), Naṭarajasana (Dancer Pose), Virabhadrasana I (Warrior I), Virabhadrasana II (Warrior II), Virabhadrasana III (Warrior III), Utthita Trikonasana (Triangle Pose), and Parivrtta Baddha Parsvakonasana (Revolved Side Angle Pose); (b) seated postures: Dasana (Staff Pose), Janu Sirsasana (Head-to-Knee Pose), Parivrtta Janu Sirsasana (Revolved Head-to-Knee Pose), Baddha Konasana (Bound Angle Pose), Gomukhasana (Cow Face Pose), Navasana (Boat Pose), and Ardha Matsyendrasana (Half Lord of the Fishes Pose); (c) knee postures: Balasana (Child’s Pose), Supta Virasana (Reclining Hero or Heroine Pose), Ustrasana (Camel Pose), Eka Pada Rajakapotasana (One-Legged King Pigeon Pose), Parighasana (Gate Pose), and Simha asana (Lion Pose); (d) prone postures: Savasana (Corpse Pose), Apanasana (Knees-to-Chest), Setu Bandha Sarvangasana (Bridge Pose), Sarvangasana (Shoulder Stand), Viparita Karani (Legs-Up-the-Wall Pose), Halasana (Plow Pose), and Jathara Parivrtti Asana (Belly Twist); and lastly, (e) supine postures: Bhujangasana (Cobra Pose), Dhanurasana (Bow Pose), and Salabhasana (Locust Pose).

All sessions were under the supervision of a professional trained in Physiotherapy with more than 5 years of experience in various styles of yoga. A criterion for analysis was set as participation in at least 80% of the scheduled sessions. Participants in the control group were recommended to continue with their daily lives without significant changes; especially they were instructed not to enroll in any training program or structured physical activity during the study period. The current levels of physical activity of the participants were controlled in both the intervention group and the control group through regular telephone calls. These calls were made periodically throughout the study period and served to collect detailed information about any physical activity in which the participants were engaged outside of the scheduled study sessions.

The general procedure for data collection included administration of questionnaires in a controlled environment, usually within the university campus, to ensure accuracy and consistency in responses. Participants completed these questionnaires under supervision to resolve any questions in real time and ensure complete understanding of each question. Regarding the schedule of evaluations, they were carried out the week before the start of the yoga intervention, thus establishing the baseline for the measurements. Subsequently, once the intervention was completed, the participants were evaluated again the following week. This scheduling of evaluations allowed immediate and effective monitoring of the effects of the intervention. Before assignment, descriptive characteristics such as age, sex, employment, physical activity, weight, height, and BMI were collected in the presence of experienced interviewers. The Body Mass Index (BMI) was calculated as weight (kg)/height (m²) (24).

Primary outcome:

Perceived stress was measured through The Perceived Stress Scale (PSS) (25) in its Spanish version (26), a 14-item questionnaire that assesses the level of stress perceived over the last month. It consists of 14 items with a response format on a five-point scale (0 = never, 1 = almost never, 2 = sometimes, 3 = often, 4 = very often). The total PSS score is obtained by reversing the scores of items 4, 5, 6, 7, 9, 10, and 13 (in the following way: 0 = 4, 1 = 3, 2 = 2, 3 = 1, and 4 = 0) and then adding up the 14 items. The direct score obtained ranges from 0 to 56 points and indicates that a higher score corresponds to a higher level of perceived stress.

Secondary outcomes:

Emotional wellbeing was measured through the Warwick-Edinburgh Mental Wellbeing Scale (WEMWBS) (27), a 14-item questionnaire on mental wellbeing that includes subjective wellbeing and psychological functioning. All items on a 5-point Likert scale are phrased positively and address aspects of positive mental health, measuring the frequency of the subject’s attitudes from “never” to “always.” Higher scores indicate better mental wellbeing. The Spanish version of WEMWBS showed good psychometric properties similar to the original scale from the United Kingdom (28).

The anxiety of the participants was measured by the State–Trait Anxiety Inventory (STAI) (29), a self-report composed of 40 items designed to assess two independent concepts of anxiety: state anxiety (a transient emotional condition) and trait anxiety (a relatively stable anxious propensity). The temporal frame of reference for state anxiety is “right now, at this moment” (20 items) and for trait anxiety is “in general, on most occasions” (20 items). Each subscale is made up of a total of 20 items in a 4-point Likert response system according to intensity (0 = almost never/none; 1 = somewhat/sometimes; 2 = moderately/often; 3 = very much/almost always). The total score on each of the subscales ranges from 0 to 60 points. Both the original questionnaire and its Spanish version have good psychometric properties, with levels of internal consistency ranging, for both the total score and each of the subscales, between 0.84 and 0.93 (30).

The sample size was determined using the freely available statistical software Epidat 4.2 (Xunta de Galicia. Consellería de Sanidade-Servizo Galego de Saúde) with the following parameters referring to the level of stress perceived as a primary outcome: it was proposed to evaluate an expected mean difference of 2.4 units between the two groups (31) or a reduction of 0.6 in each group (32). With a significance level of 5% (corresponding to a z-value of 1.96) and a power of 90% (corresponding to a z-value of 1.28). A loss percentage of 15% was also considered, which allowed adjusting the sample size to 122 participants to compensate for possible dropouts or missing data during the study.

For population characterization, categorical variables were presented in terms of frequency and percentage. An exploratory analysis was conducted to validate the normal distribution of quantitative variables, ensuring adherence to the requisite assumptions for each analysis (Kolmogorov–Smirnov >0.05). Subsequently, mean values and standard deviations were computed for each variable of interest. The Student’s t-test for independent samples was employed to assess inter-group differences.

Moreover, an effective treatment analysis was executed. This entailed the utilization of a mixed analysis of variance, with the intervention (YGE vs. CG) acting as the between-groups factor and the measurement time (pre-treatment versus post-treatment) serving as the within-subject variable. Dependent variables, such as perceived stress, emotional wellbeing, and anxiety, underwent separate analyses. An exploration of potential interactions between treatment group and measurement time was conducted.

Additionally, Cohen’s d was utilized to determine intergroup effect sizes, with values of ≤0.2, 0.5, and 0.8 indicating small, medium, and large effects, respectively. A significance level of p < 0.05 was established for all analyses. Bonferroni correction was used to adjust the p-values for independent samples t-tests and for multiple comparisons in the case of the mixed analysis of variance. Statistical computations were performed using SPSS statistical software, version 17.0 (SPSS, Inc., Chicago, IL, United States).

Before the intervention, participants in the Experimental Group (EG) reported higher values in the primary variable of perceived stress (25.82 ± 7.24) compared to those in the Control Group (CG) (24.83 ± 6.25). However, post-measurement revealed higher stress levels in the CG (25.09 ± 6.41) compared to the EG (22.77 ± 6.45). Significant differences were found in Group × Time: F(1, 127) = 16.853, p = 0.001, η² = 0.117, and in Time: F(1, 127) = 11.880, p = 0.001, η² = 0.086, but not in Group: F(1, 127) = 0.376, p = 0.541, η² = 0.003 (Figure 2). The comprehensive analysis showed statistically significant differences between groups in the post-intervention measurement, t(127) = 2.053, p = 0.042, with a small effect size (d = 0.36). Additionally, significant differences were observed in pre and post-measurements for the group undergoing yoga treatment/training, t(64) = 3.921, p = 0.001, with a small effect size (d = 0.44). After Bonferroni correction, the differences remained significant: t(127) = 2.053, p = 0.042* (adjusted), d = 0.36 (small effect size) for group comparison, and t(64) = 3.921, p = 0.001* (adjusted), d = 0.44 (small effect size) for pre-post comparisons. (*Significant after Bonferroni correction, p < 0.05/3 = 0.017) (Table 2).

In emotional wellbeing, participants in the Control Group (CG) initially reported higher values (44.25 ± 7.46) compared to those in the Experimental Group (EG) (43.42 ± 8.20). However, conversely, the EG participants (47.00 ± 7.10) achieved higher values than the CG (44.06 ± 7.55). Significant differences were observed in Group × Time: F(1, 127) = 11.383, p = 0.001, η² = 0.082, and in Time: F(1, 127) = 9.232, p = 0.003, η² = 0.068, but not in Group: F(1, 127) = 0.751, p = 0.388, η² = 0.006 (Figure 3). After Bonferroni correction, the comprehensive analysis revealed statistically significant differences between both groups in the post-intervention measurement, t(127) = −2.277, p = 0.024, with a small effect size (d = 0.08). Additionally, statistically significant differences were observed between the pre and post-measurements in the group that received treatment/training in yoga, t(64) = −3.405, p = 0.001, with a small effect size (d = 0.47) (Table 2).

Regarding state anxiety, participants in the Control Group (CG) reported higher values (37.39 ± 12.56) than those in the Experimental Group (EG) (32.69 ± 12.12) before the start of the intervention, as well as in the post-intervention measurement (37.05 ± 12.75 vs. 28.42 ± 10.26), and significant differences appeared in Group × Time: F(1, 127) = 8.008, p = 0.005, η² = 0.059, in Time: F(1, 127) = 11.053, p = 0.001, η² = 0.080, and in Group: F(1,127) = 11.246, p = 0.001, η² = 0.000 (Figure 4). The comprehensive analysis of the interaction demonstrates the existence of statistically significant differences between both groups in the post-intervention measurement, t(127) = 4.240, p = 0.001* (adjusted with Bonferroni correction), with a small effect size (d = 0.75). Additionally, the existence of statistically significant differences between the pre and post-measurement in the group that received the treatment/training in yoga was observed, t(64) = 3.148, p = 0.002* (adjusted with Bonferroni correction), with a medium effect size (d = 0.38).

Regarding trait anxiety, participants in the Experimental Group (EG) reported higher values (40.74 ± 12.37) than those in the Control Group (CG) (37.34 ± 10.33) before the start of the intervention. Conversely, in the post-measurement, the CG obtained higher values (36.19 ± 10.38) than the EG (31.95 ± 9.56), and significant differences appeared in Group × Time: F(1.27) = 16.634, p = 0.001, η² = 0.116 and in Time: F(1,127) = 24.248, p = 0.000, η² = 0.182, but not in Group: F(1, 127) = 0.066, p = 0.798, η² = 0.001 (Figure 4). The comprehensive analysis of the interaction demonstrates the existence of statistically significant differences between both groups in the post-intervention measurement, t(102) = 2.410, p = 0.017* (adjusted with Bonferroni correction), with a small effect size (d = 0.42). Additionally, the existence of statistically significant differences between the pre and post-measurement in the group that received the treatment/training in yoga was observed, t(51) = 6.406, p = 0.001* (adjusted with Bonferroni correction), with a medium effect size (d = 0.80) (Table 2; Figure 5).