Estimating the effect of exercise on mental health : review and Bayesian network meta-analysis

Mental health disorders are debilitating and costly. Although effective pharmacological and psychological treatments exist, first-line treatments fall short for a meaningful proportion of individuals. Further, barriers such as economic limitations or unfavorable attitudes toward treatments impede receipt of services. Exercise for mental health has received particular attention in recent decades as an accessible, acceptable, and cost-effective intervention for mental health disorders. This dissertation aims to contribute to this literature by providing an overview and by applying an advanced meta-analytic technique to summarize the effects of exercise for depression. First, I conduct a qualitative overview of the literature of exercise for mental health. This includes the acute effects of exercise on affect and cognition, and long-term effects from cross-sectional and longitudinal studies. Next, I summarize the evidence for putative mechanisms that have received particular attention, including psychosocial constructs (e.g., general self-efficacy) and neurobiological elements (e.g., brain-derived neurotrophic factor). I synthesize the available clinical trials of exercise as an intervention for any mental health disorders. This review of the literature reveals that an impressive volume of empirical evidence has been collected recently, but also highlights inconsistency. In major depressive disorder (MDD), the mental health diagnosis with the largest body of evidence for exercise, a number of meta-analyses have been conducted with a broad range of conclusions. I discuss the rationale and methodology of network meta-analysis as an approach to resolve these conflicts, including the advantages, assumptions, statistics, graphical presentations, and interpretation. Finally, I conduct a Bayesian network meta-analysis of exercise for MDD. The overall effect of exercise was significant and large compared to waitlist or no treatment (SMD = 1.04) and significant and medium compared to inpatient treatment as usual (SMD = 0.57). The overall effect of exercise did not significantly differ from antidepressant medication (ADM) or cognitive-behavior therapy (CBT). The effect of adjuvant exercise showed a small-to-medium, nonsignificant effect when added to ADM or CBT compared to those interventions alone, indicating that more evidence is needed. In a second model, effects of specific interventions (e.g., resistance exercise; aerobic exercise + sertraline) are estimated. In general, confidence intervals were quite broad, suggesting that more evidence is needed and the size of effects should be interpreted with caution. To conclude, the best available estimate of the effect of exercise for MDD indicates that it is similar to ADM or CBT in its effects, and may exert a small additional effect when added to monotherapy. While effects of specific types of interventions may vary, nearly all exercise approaches outperformed waitlist, including nonspecific physical activity interventions. Finally, the significant effect of exercise compared to psychiatric TAU, along with the evidence that inpatient psychiatric TAU has a small effect, suggested that supervised exercise should be considered as an effective addition to inpatient treatment.