Testing the Penn Resiliency Program's Theoretical Model in a Randomized Controlled Trial

Background: The Penn Resiliency Program (PRP), which aims to prevent depression in adolescents, has established efficacy in reducing depressive symptoms (Brunwasser et al., 2009).

Objective: To test the plausibility of the theoretical model of change for PRP, a cognitive-behavioral (CB) depression prevention program for adolescents.

Method: Middle school students (N = 697) were randomized to one of three conditions; PRP; an assessment-only control condition (CON); or the Penn Enhancement Program (PEP), which is a placebo-control condition. In the primary outcome paper, Gillham and colleagues (2007) found that PRP reduced depressive symptoms relative to both CON and PEP in two of three participating schools (A & B); however, in school C, PEP reduced depressive symptoms more than PRP. We evaluated explanatory style and depressive symptoms across six time points spanning 24 months of follow-up. A two-stage piecewise parallel process growth model was used to test a moderated-mediation hypothesis: Within schools A & B but not C, PRP’s effects on 24-month levels of depressive symptoms were indirect through initial improvements in explanatory style from baseline through the 12-month follow-up. In the same model, we tested an alternative mediation pathway in which PRP’s effect on 24-month levels of explanatory style was indirect through initial improvements in depressive symptoms (i.e., reversing the hypothesized direction of causality). Confidence intervals for indirect effects were calculated using the bias-corrected bootstrap method with 5000 draws.

Results: Relative to CON, there was a significant indirect effect of PRP on 24-month levels of depressive symptoms through improvements in explanatory style within schools A & B: γ = -0.17, 95% CI [0.03, 0.47]. Relative to PEP, the indirect effect showed a nonsignificant trend:  γ = 0.13, 95% CI [-0.004, 0.41]. There was no evidence for the reverse-causality hypothesis: Indirect effects of PRP relative to both PEP and CON on 24-month levels of explanatory style through initial improvements in depressive symptoms were not significant in either schools A & B combined or in school C. There was no evidence of group differences in growth in explanatory style within school C.

Conclusions: These findings are generally consistent with the CB theory upon which PRP was founded. In the schools where PRP was effective, improvements in explanatory style was a mediator of the effect on depressive symptoms. Future studies should evaluate competing mediators of PRP’s effects and identify moderators that may account for PRP’s inconsistent performance across schools.