What meta-analyses have and have not taught us about psychotherapy effects: A review and future directions

Sixty-three meta-analyses of psychotherapeutic interventions are reviewed to examine what we have learned from them about the causal nature, magnitude and generalizability of psychotherapy effects. Despite the large number of meta-analyses demonstrating specific and nonspecific causal effects of psychotherapeutic interventions, limitations in outcome studies and in meta-analytic reviews currently prevent us from drawing strong generalized inferences about the magnitude of effects, conditions that moderate their size, and variables that mediate therapy effects. There is mounting evidence suggesting that biases associated with individual studies do not cancel each other out when studies are combined meta-analytically leading to inflated mean effect estimates for some interventions, diagnoses, and outcomes, and deflated estimates for others. There is currently little agreement in the reviewed meta-analyses concerning the limits of the apparent robustness of intervention effects and variables considered as potential moderators of psychotherapy effects are frequently confounded with other study characteristics, Moreover; generalizability of psychotherapy effects to major target universes (e.g., clinical practice settings, African-American and Latino populations) is compromised by the predominance of efficacy studies conducted in controlled research settings on Angle populations. Future directions for meta-analyses of psychotherapy outcome studies are discussed. (C) 1997 Elsevier Science Ltd.