Longitudinal effects of compassion training on emotion regulation and prosocial moral development among adolescents

This cluster-randomized trial compared eight weeks of compassion training with an intensity-matched relaxation program in 876 adolescents from 34 classes in four Chinese secondary schools. Assessments occurred at baseline, post-intervention, and 3- and 12-month follow-ups. Three-level models showed greater baseline-to-follow-up change under compassion training at all post-baseline waves for emotion regulation (ER) and prosocial moral development (PMD; all p < 0.001). At 12 months, differential change was 2.203 points for ER (approximate 95% CI 1.266-3.140; baseline-SD-standardized effect 0.317) and 2.649 points for PMD (1.526-3.772; standardized effect 0.278). Piecewise models showed larger intervention-period gains followed by partial attenuation, supporting H1 and H2 and partially supporting H3. In the compassion condition, ER predicted subsequent PMD across all three intervals, whereas the reverse paths were significant in the first two intervals only. Between-condition Wald tests were nonsignificant for all six cross-lagged parameters (p = 0.067-0.648), providing no support for H4. Retention at 12 months was 59.6%. Differential changes remained positive and significant across delta-adjusted sensitivity scenarios, although their magnitudes decreased as departures from MAR increased. Compassion training was associated with sustained relative advantages in this sample, without evidence of stronger reciprocal coupling than relaxation training.