This study commenced in May and was completed in its entirety on August 25, at a comprehensive university in Sichuan Province, encompassing the following phases: protocol preparation, participant recruitment and informed consent, intervention implementation, questionnaire/scale assessments, and data collection. The university is well-equipped with sports facilities, exercise venues, and professional sports instructors, meeting the basic requirements for the trial. Participants were recruited using convenience sampling.

PTSD symptom status was operationalized using the PCL-5. Participants were not clinically diagnosed with PTSD for the purposes of this study; instead, eligibility was based on a screening cutoff, which indicates elevated or probable PTSD symptomatology in prior research. Accordingly, we refer to the sample as college students with elevated PTSD symptoms rather than PTSD patients.

The inclusion criteria were: (1) Full-time college students aged 18–25; (2) A preliminary screening for PTSD using the PTSD Checklist 5 (PCL-5) (score ≥ 53, indicating PTSD symptoms); (3) Voluntary participation in the study and signing of an informed consent form; (4) No severe cognitive impairments and sufficient proficiency in Chinese for effective communication; (5) Physical condition permitting normal exercise.

Exclusion criteria were: (1) Severe exercise injuries or illnesses occurring during the study; (2) Failure to complete pre-and post-intervention assessments; (3) Current engagement in other systematic psychological or pharmacological treatments.

The minimum sample size was calculated using G*Power software (3.1). This study employed paired sample t-tests with a significance level (α) set at 0.05 and an effect size set at 0.5. To achieve a statistical power of 80% (1-β), at least 34 participants were needed. Considering a dropout rate of 15%, the study planned to recruit no fewer than 40 participants.

The study received approval from the Academic Ethics Committee of North Sichuan Medical College (No.:2025020). All participants were informed of the activity’s purpose, procedures, and reward mechanisms, and signed written informed consent before the study began. They were informed that they could withdraw at any time unconditionally. The research process strictly adhered to confidentiality principles, and data were processed anonymously.

The pre-test was conducted before the intervention began, and the post-test was completed within 1 week after the 21-day activity ended. The intervention scheme adopted a campus activity format, implementing a “21-Day Exercise Check-In Activity” lasting 3 weeks. Participants were required to choose from three types of exercise: (1) Jump rope for ≥10 min (daily submission of jump rope videos); (2) Run ≥2 kilometers (daily submission of screenshots from the Keep App); (3) Complete a full set of Tai Chi movements (daily submission of a full-length video). All submitted materials (videos or screenshots) must display time information, and forgery is prohibited. Researchers and campus activity staff conducted random on-site checks to ensure authenticity. Participants had to complete 21 consecutive days to qualify for inclusion in the study. Given the relatively large amount of exercise in 30 min, the study required only 30 min of exercise over a single day, and the duration of each exercise session was recorded, amounting to 30 min over a single day.

Participants who completed all assessments (pre-test, post-test) and maintained ≥ 80% exercise session adherence received a gift card valued at approximately ¥50 (approximately $7 USD) for a popular online shopping platform.

Data included the results from pre- and post-assessment measures and the 21-day check-in records, all collected and coded by the research team. Statistical analyses were performed using SPSS 26.0. Descriptive statistics were used to analyze demographic characteristics. For differences between pre- and post-assessments, paired sample t-tests were applied if the data met normality assumptions; otherwise, Wilcoxon signed-rank tests were used.

Effect size estimation and confidence intervals. For paired-samples t tests, we reported Cohen’s d_z as the standardized mean change, calculated as d_z = t/√N (equivalently, mean of the paired differences divided by the SD of the paired differences). For Wilcoxon signed-rank tests, we reported the effect size r = Z/√N, where N is the number of non-zero paired differences. Ninety-five percent confidence intervals (CIs) for r were estimated using percentile bootstrap resampling (5,000 iterations) of paired observations (no continuity correction). In cases where the bootstrap distribution showed negligible variability, the percentile CI could collapse to a single value.

Change-score (within-person) analytic strategy. Given the single-group pre–post design, observations are not independent if pre- and post-test records are stacked at the measurement level. PROCESS Model 4 is a single-level ordinary least squares (OLS) mediation procedure and does not inherently account for within-subject clustering. Therefore, to avoid violating the independence assumption and to provide a transparent exploratory test of the hypothesized mechanism, we conducted analyses using within-person change scores. For each participant, we computed change scores as Δ = post -pre for PTSD, anxiety, depression, and psychological resilience. We then examined whether changes in psychological resilience were associated with changes in PTSD/anxiety/depression using linear regression models (with 95% confidence intervals).

Given the exploratory nature of these analyses and the absence of a control group, results should be interpreted as preliminary evidence of potential mechanisms rather than confirmatory tests of causal mediation. The statistical significance level was set at p < 0.05.

Attrition from baseline (n = 76) to the final analytic sample was 38.2% (n = 47). Reasons for post-baseline attrition included reported physical discomfort during the intervention (n = 15), voluntary withdrawal (n = 7), and missing post-test questionnaires (n = 7). Because attrition may not be random, the observed pre–post changes should be interpreted cautiously as they may over- or under-estimate symptom changes in the initially recruited population.

Of the 36 non-completers, 15 participants withdrew specifically due to adverse physical symptoms during the intervention sessions. Reported symptoms included dizziness (n = 8), muscle soreness (n = 5), and combined symptoms (n = 2). These symptom-related dropouts occurred predominantly during the initial intervention period (Week 1: n = 9; Week 2: n = 4; Week 3: n = 2), suggesting that early exercise sessions posed the greatest challenge for vulnerable participants.

Informal post-withdrawal contact with these participants revealed that symptoms were typically experienced during or immediately after higher-intensity aerobic segments. Importantly, no participants required medical attention, and all reported symptom resolution within 24–48 h of discontinuation. Nevertheless, the concentration of symptom-related dropout among participants with elevated baseline psychopathology raises concerns about the tolerability of the exercise protocol for more severely affected individuals.

We calculated the differences in scores for PTSD, anxiety, depression, and psychological resilience for both pre-test and post-test, followed by exploratory analysis to test for normality. Results indicated that anxiety (p = 0.295) and depression (p = 0.051) exhibited significant normality, while psychological resilience (p = 0.026) and PTSD (p < 0.001) showed non-normality. Thus, paired sample t-tests were used for anxiety and depression, whereas Wilcoxon signed-rank tests were applied for psychological resilience and PTSD.

We conducted independent samples t-tests to compare baseline characteristics between participants who completed the study (n = 47) and those who dropped out (n = 36).

As shown in Table 3, significant differences emerged between completers and non-completers on all baseline measures (all p < 0.001). Notably, participants who dropped out exhibited significantly higher levels of PTSD symptoms, anxiety, and depression, as well as significantly lower psychological resilience, compared to those who completed the intervention. These effect sizes range from medium to large, indicating substantial and clinically meaningful baseline differences between the two groups.

These findings have important implications for interpreting the study results. The pattern suggests that participants with more severe symptomatology and lower resilience resources were disproportionately likely to discontinue participation. This differential attrition may have resulted in a final sample that was less symptomatic and more psychologically resilient than the initially enrolled population, potentially inflating observed treatment effects and limiting generalizability to more severely affected individuals.

To further examine the intervention effects of school-based regular aerobic exercise intervention on PTSD, anxiety, depression, and psychological resilience, paired sample t-tests were conducted on anxiety and depression scores pre- (T1) and post-intervention (T2) for all participants, while Wilcoxon signed-rank tests were used for psychological resilience and PTSD. Effect sizes were measured using Cohen’s d, with Cohen’s (1988) standards indicating d = 0.2 as a small effect, d = 0.5 as a medium effect, and d = 0.8 as a large effect, as detailed in Table 4.

The paired sample t-test results showed that pre-test anxiety scores (M = 3.489, SD = 0.718) were significantly higher than post-test anxiety scores (M = 2.648, SD = 0.736). This indicates that school-based regular aerobic exercise intervention has a significant impact on reducing anxiety symptoms among college students (t = −6.352, p < 0.001, Cohen’s d = −0.926, 95% CI = [−1.266, −0.580]). The effect size far exceeds the threshold of 0.8, indicating a very large effect of the intervention on improving anxiety symptoms.

Pre-test depression scores (M = 3.514, SD = 0.647) were significantly higher than post-test depression scores (M = 3.087, SD = 0.383). This suggests that school-based regular aerobic exercise intervention significantly impacts depression symptoms in college students (t = −3.963, p < 0.001, Cohen’s d = −0.578, 95% CI = [−0.885, −0.266]). The effect size falls within the medium range, suggesting a notable improvement effect on depression symptoms.

Wilcoxon signed-rank test results indicated that pre-test psychological resilience scores (M = 2.843, SD = 0.691) were significantly lower than post-test psychological resilience scores (M = 3.585, SD = 0.534). This indicates that school-based regular aerobic exercise intervention has a significant effect on enhancing psychological resilience among college students (Z = 4.842, p < 0.001, r = 0.707, 95% CI = [0.544, 0.814]). Pre-test PTSD scores (M = 3.807, SD = 0.749) were significantly higher than post-test scores (M = 2.404, SD = 0.855), indicating that school-based regular aerobic exercise intervention can effectively improve PTSD symptoms (Z = −5.975, p < 0.001, r = 0.871, 95% CI = [−0.871,–0.871]).

To complement the statistical significance findings, we examined the clinical meaningfulness of observed changes by analyzing categorical shifts in symptom severity and the proportion of participants crossing established clinical thresholds.

These exploratory analyses suggest that changes in psychological resilience may be associated with changes in anxiety and PTSD symptoms following the exercise intervention. Specifically, participants who experienced greater enhancement in resilience also tended to experience greater symptom relief.

First, due to the single-group design, we cannot determine whether the intervention caused the observed changes or whether these changes would have occurred naturally over time. Second, the change-score correlations do not establish temporal precedence—we cannot determine whether resilience changes preceded symptom changes or vice versa. Third, the modest effect sizes (R² ranging from 0.05 to 0.11) indicate that resilience changes account for only a small proportion of the variance in symptom changes, suggesting that other unmeasured factors likely play important roles.

These findings should therefore be considered hypothesis-generating rather than confirmatory and warrant replication in randomized controlled trials with appropriate temporal measurement of mediators and outcomes.

The results of the single group pre-and post-test analysis indicate that regular aerobic exercise significantly promotes the psychological health of participants. Engaging in school-based regular aerobic exercise intervention provides college students an immediate sense of control, with physiological responses (like adjusted breathing and increased heart rate) and pleasurable experiences serving as positive feedback (56). This accumulation of positive experiences continually strengthens the individual’s motivation to exercise and adherence to the behavior. Consequently, it effectively breaks the passive and helpless feelings that trauma survivors may develop in daily life, helping them reconstruct a positive self-schema through specific exercise practices.

Regular exercise tasks require students to complete activities at fixed times or in specific contexts, thus providing a structured coherence to their lives (57). For college students experiencing psychological distress, having a structured time and behavior pattern significantly reduces the intrusion of random stressors and avoids the psychological emptiness that may arise from a lack of goals and planning. More critically, this rhythmic arrangement enables students to gradually experience the transition from chaos to order, enhancing their sense of stability and security. Security buffers subjective feelings of external threat and diminishes excessive vigilance toward the environment (58).

It is noteworthy that regular aerobic exercise also serves as a significant opportunity for emotional transformation. After completing exercise, individuals typically experience feelings of joy, relaxation, and even a slight sense of accomplishment, introducing new variables into previously rigid negative emotional patterns (59). Individuals under constant stress often tend to interpret external events negatively, leading to cognitive rigidity (60). Exercise provides a means to interrupt this pattern, allowing them to gain psychological relief from brief emotional peaks (61). As these positive experiences accumulate, individuals’ cognitive flexibility gradually increases, manifested through more open thinking, adaptive self-talk, and positive meaning construction.

The present study demonstrated not only statistically significant improvements across all three primary outcome measures, but also clinically meaningful changes that underscore the therapeutic value of the online group intervention for earthquake-affected adolescents.

Regarding PTSD symptomatology, the intervention produced substantial clinical benefits. At baseline, all 47 participants met the clinical threshold for probable PTSD, reflecting the severe psychological impact of earthquake trauma on this population. Following the intervention, 27 participants (57.45%) transitioned below the clinical cutoff, indicating diagnostic-level remission. Furthermore, reliable change index analysis revealed that 41 participants (87.23%) exceeded the threshold for reliable improvement (RCI > 1.96), demonstrating that the observed reductions in PTSD symptoms were not attributable to measurement error. The mean reduction of 18.36 points on the PCL-5 substantially exceeded the established minimal clinically important difference of 10–20 points, suggesting that participants experienced perceptible and meaningful relief from intrusive memories, avoidance behaviors, negative cognitions, and hyperarousal symptoms.

For anxiety symptoms, the clinical impact was equally pronounced. Among the 41 participants who met criteria for clinical anxiety at baseline, 25 (60.98%) achieved remission to non-clinical status following the intervention. Category-level analysis further demonstrated that 36 participants (76.60%) improved by at least one severity category, with only 3 participants (6.38%) showing category-level worsening. Notably, of the 14 participants classified in the moderate–severe or severe anxiety ranges at baseline, 13 (92.86%) improved to less severe categories, with the number of severe cases decreasing from 6 to zero. These patterns indicate that the intervention was particularly effective for those with more pronounced anxiety presentations.

Depression outcomes similarly reflected meaningful clinical improvement, though the pattern differed somewhat from the other measures. Of the 44 participants meeting clinical criteria for depression at baseline, 11 (25.00%) transitioned to the non-clinical range. While this remission rate was lower than those observed for PTSD and anxiety, category-level analysis revealed that 25 participants (53.19%) improved by at least one severity category, with only 2 (4.26%) showing worsening. Particularly noteworthy was the complete resolution of severe depression cases: all 19 participants who were in the moderate–severe or severe depression categories at baseline improved to less severe categories, with no participants remaining at these highest severity levels post-intervention.

Collectively, these findings indicate that the online group intervention produced clinically meaningful benefits across the spectrum of trauma-related psychopathology. The convergence of multiple indicators—including remission rates, reliable change indices, and severity category transitions—provides robust evidence that the observed statistical effects translated into tangible improvements in participants’ psychological functioning. The particularly strong effects observed for PTSD symptoms align with the trauma-focused nature of the intervention, while the substantial improvements in anxiety and depression suggest beneficial effects on the broader constellation of post-disaster psychological distress.

The exploratory change-score analyses provide preliminary evidence suggesting that improvements in psychological resilience may be associated with reductions in anxiety and PTSD symptoms following the exercise intervention. Specifically, participants who showed greater increases in resilience also tended to show greater decreases in anxiety (β = −0.335, p = 0.021) and PTSD symptoms (β = −0.329, p = 0.024). These associations, while modest in magnitude, are consistent with theoretical frameworks positing resilience as a dynamic psychological resource that influences trauma recovery (62).

However, it is crucial to interpret these findings within the constraints of our study design. The single-group pre-post design precludes causal inference, as we cannot rule out alternative explanations for the observed associations. The correlational nature of change-score analyses means that we cannot determine temporal precedence—whether resilience improvements preceded symptom relief, occurred simultaneously, or followed symptom changes. Additionally, third variables not measured in this study may account for the observed associations between resilience and symptom changes.

The non-significant association between resilience change and depression change (β = −0.231, p = 0.118) warrants consideration. While the direction of the effect was consistent with hypotheses, the lack of statistical significance may reflect insufficient statistical power given our sample size, different mechanisms underlying depression versus anxiety and PTSD, or the possibility that resilience operates differently on depression within the context of exercise interventions. Future studies with larger samples and more frequent measurement occasions could help clarify these relationships.

Despite these limitations, the findings provide a foundation for hypothesis development regarding the mechanisms through which exercise interventions may affect psychological outcomes. The consistent pattern of associations between resilience enhancement and symptom reduction across anxiety and PTSD outcomes suggests that resilience-building components may be valuable targets for optimization in future intervention designs. These findings align with conservation of resources theory (Alvaro et al., 2010), which suggests that resource acquisition through exercise may buffer against the resource depletion associated with trauma exposure.

These findings carry several implications for practice. First, schools should consider integrating brief, structured exercise opportunities into daily routines as a component of mental health promotion. The online delivery format used in this study demonstrates feasibility for implementation without requiring specialized facilities or extensive personnel training. Schools might incorporate guided exercise sessions into morning routines, physical education classes, or after-school programming, with particular attention to students showing elevated depressive symptoms.

Second, the accessibility of the exercise modalities studied—jump rope, running, and Tai Chi—suggests potential for broad implementation. These activities require minimal equipment, can be performed in limited spaces, and are adaptable to varying fitness levels. The flexibility to choose and switch between modalities enhanced acceptability in our sample and may be an important design feature for real-world programs. Mandating a single exercise type may be less effective than offering structured choice within defined parameters.

Third, the online platform approach offers scalability advantages, particularly for reaching adolescents in under-resourced areas or those who face barriers to in-person programming. However, our attrition rate indicates that online delivery alone is insufficient; the combination of digital content with human support (facilitator check-ins, personalized feedback) was likely important for maintaining engagement. Future implementations should preserve this blended approach rather than relying solely on automated content delivery.

Fourth, the 21-day duration, while sufficient to produce measurable effects, is brief relative to the chronic nature of adolescent depression. Programs should plan for maintenance and follow-up support to sustain initial gains. This might include transitioning participants to ongoing exercise opportunities within schools or communities, periodic booster sessions, or peer support networks that encourage continued physical activity.

Finally, exercise interventions should be positioned as complementary to, rather than substitutes for, established mental health treatments. Adolescents with moderate-to-severe depression or active suicidal ideation require professional clinical care; exercise may serve as an adjunct that enhances treatment response or as a preventive strategy for those with subclinical symptoms.

The withdrawal of 15 participants due to adverse physical symptoms represents a significant concern that warrants careful examination. When considered alongside the finding that non-completers exhibited significantly higher baseline PTSD, anxiety, and depression scores, a troubling pattern emerges: the intervention may be least tolerable for those who are most symptomatic and potentially most in need of treatment.

Several mechanisms may explain the elevated vulnerability of more symptomatic participants to exercise-induced adverse effects. First, individuals with elevated PTSD and anxiety often exhibit chronic hyperarousal, including elevated resting heart rate, heightened sympathetic nervous system activation, and dysregulated stress response systems. The additional physiological demands of exercise may have compounded this existing arousal, potentially triggering symptoms such as dizziness, lightheadedness, or panic-like sensations. Second, trauma-affected individuals frequently demonstrate altered interoceptive processing, with some experiencing hypervigilance to bodily sensations that may amplify the perceived intensity of normal exercise-related experiences (e.g., increased heart rate, sweating, breathlessness) and trigger trauma-associated fear responses. Third, elevated depression is associated with physical deconditioning, fatigue, and reduced exercise tolerance, potentially rendering standard exercise intensities more challenging for these individuals. Fourth, lower psychological resilience may reduce capacity to persist through physical discomfort that more resilient individuals might tolerate.

The concentration of symptom-related dropouts in the first 2 weeks suggests that the initial adaptation period poses the greatest risk. This finding has direct implications for protocol design: more gradual introduction of exercise, with extended low-intensity acclimation periods, may be essential for more symptomatic participants.

These findings do not necessarily contraindicate exercise-based interventions for trauma-affected populations. Indeed, substantial evidence supports the benefits of exercise for PTSD and related conditions. However, our results suggest that standard exercise protocols may require significant modification for individuals with more severe symptomatology. Recommendations for future implementation include: Comprehensive pre-intervention assessment of current fitness levels, exercise history, physical health conditions, and symptom severity to identify individuals at elevated risk for adverse responses; Development of differentiated exercise prescriptions based on baseline symptom severity and fitness levels, with lower-intensity starting points for more affected individuals; Longer initial phases with very gradual intensity progression, particularly for participants with elevated symptoms or limited exercise experience.

Several directions for future research emerge from this study’s findings and limitations. Most urgently, randomized controlled trials with longer follow-up periods are needed to establish whether observed effects persist beyond the immediate post-intervention period and to identify optimal intervention duration for sustained benefit. Our 21-day protocol produced meaningful short-term effects, but the durability of these effects remains unknown.

Future studies should incorporate objective measures of both exercise behavior (e.g., accelerometry) and biological outcomes (e.g., inflammatory markers, cortisol, neuroimaging) to reduce reliance on self-report and illuminate mechanisms of action. Such measures would address the methodological limitations of the present study and provide stronger evidence for causal claims.

Research should examine potential moderators of intervention response. Our sample was insufficiently powered to detect differential effects by sex, baseline symptom severity, exercise history, or exercise modality preference. Identifying which adolescents benefit most—and which may require modified or alternative approaches—would enable more efficient resource allocation.

Implementation research examining real-world adoption, adaptation, and sustainment of exercise-based mental health programming in schools and communities would bridge the gap between controlled research and practical impact. Identifying barriers and facilitators to implementation, cost-effectiveness relative to other mental health interventions, and strategies for maintaining program quality at scale represent important next steps toward translating this evidence into public health benefit.

This study primarily recruited participants from the same university, resulting in a concentration of cultural backgrounds, socioeconomic levels, and campus environments; thus, the generalizability and cross-cultural applicability of the findings still require validation. This bias limits the external validity of the results, which might overestimate the true group effects. Future research should strive for diversity in samples and breakthroughs in cross-cultural effects. Expanding the intervention to adolescents in middle school, workplace populations, or even veterans can allow for comparisons of exercise effects in different developmental tasks and social pressure contexts.

Attrition and selection bias. A substantial proportion of participants did not complete the intervention and post-test assessments. This level of attrition raises concerns about selection bias and limits generalizability. Participants who completed the 21-day check-in may have differed systematically from those who discontinued (e.g., greater exercise tolerance, higher motivation, fewer time constraints, or different baseline symptom severity). Consequently, the magnitude of within-person symptom change observed among completers may not represent the change that would occur in the broader screened population. Future studies should implement strategies to reduce attrition (e.g., graded intensity options, adverse-event monitoring and management, flexible scheduling, and enhanced follow-up procedures) and should examine baseline differences between completers and non-completers, as well as apply analytic approaches that better accommodate missing data (e.g., mixed-effects models and/or multiple imputation), to strengthen the robustness of conclusions.

Importantly, the present findings should be interpreted in light of the single-group pre-post design. Symptom reductions and resilience gains may reflect multiple alternative explanations, including regression to the mean, spontaneous symptom fluctuation and natural recovery over time, repeated-measurement effects, seasonal or contextual changes during the study period, or concurrent informal support. Therefore, the results provide preliminary evidence of feasibility and within-person change, but they do not demonstrate causal effectiveness. Future studies should employ randomized controlled designs (e.g., waitlist or active control conditions), objective activity monitoring (e.g., wearables), and longer-term follow-up to test durability and to strengthen causal inference regarding mechanisms.

The substantial attrition rate and the systematic differences between completers and non-completers represent significant threats to internal and external validity. Of the 83 participants with complete baseline data, 36 (43.37%) did not complete the intervention. Independent samples t-tests revealed that non-completers exhibited significantly higher baseline levels of PTSD, anxiety, and depression, as well as significantly lower psychological resilience, compared to completers. This pattern of differential attrition indicates that the intervention may have been less acceptable, tolerable, or accessible to those with more severe psychopathology and fewer psychological resources—precisely the individuals who may be in greatest need of intervention. The implications of this attrition pattern are substantial. The final analytic sample likely represents a more resilient, less symptomatic subgroup of the initially enrolled population, potentially inflating effect size estimates and remission rates. The true efficacy of the intervention for adolescents with more severe trauma-related symptoms remains uncertain. Clinicians and researchers considering implementation of similar interventions should be aware that individuals presenting with higher symptom severity and lower resilience may require additional support, modified protocols, or alternative treatment approaches to enhance retention and benefit from exercise-based interventions.

Beyond the measured baseline differences, self-selection processes may have introduced additional unmeasured confounds. Participants who volunteered for an exercise-based intervention and persisted through completion may have possessed more positive attitudes toward physical activity, greater self-efficacy for behavior change, stronger family support, or other characteristics associated with favorable treatment response. These unmeasured factors may further limit generalizability to the broader population of earthquake-affected adolescents.

Due to the single-group pre-post design, we employed change-score regression analyses. While this approach is transparent and appropriate for our design, it cannot establish that resilience changes causally mediate intervention effects on symptoms. The change-score correlations observed may reflect shared method variance, common causes affecting both resilience and symptom changes, or reverse causation (symptom improvement leading to resilience enhancement rather than vice versa). Future research employing randomized controlled trials with multiple measurement waves would enable more rigorous mediation analysis using appropriate statistical techniques such as latent change score models or multilevel structural equation modeling, which can better account for measurement error and establish temporal precedence of mediators relative to outcomes.