Exercise as a treatment for depression: A meta-analysis adjusting for publication bias
Introduction
Depression is a prevalent condition, with a long-life prevalence ranging from 10% to about 20% in different countries (Andrade et al., 2003). Depression is a major cause of disability, responsible for 40.5% of total disability-adjusted life years (DALYs) caused by mental and substance-use disorders (Whiteford et al., 2013).
Physical activity and exercise are suggested as potential treatments for depression, and incorporated in guidelines as a complementary form for illness of mild to moderate severity (Cleare et al., 2015). Several meta-analyses have demonstrated that exercise is an effective treatment for depression, with a pooled standardized mean deviation (SMD) ranging from small (−0.4) (Krogh et al., 2011) to very large (−1.4) (Cooney et al., 2013, Craft and Landers, 1998, Daley, 2008, Danielsson et al., 2013, Josefsson et al., 2014, Krogh et al., 2011, Rethorst et al., 2009, Silveira et al., 2013, Stathopoulou et al., 2006). However, a number of different approaches have been undertaken in prior meta-analyses and uncertainty remains over the magnitude of the effects of exercise on depression.
The 2013 update of the Cochrane review on exercise for depression, provided new data for discussion, showing that when analysis was restricted to the six trials considered of low risk of bias only, the SMD was small and non-significant (Cooney et al., 2013). This review has been criticized, with a particular emphasis on the potential inappropriate selection criteria applied (Ekkekakis, 2015). For example, the review proposed excluding studies that had a control arm with any “active control comparison”. However, some studies that compared different exercise arms were included (Krogh et al., 2009), thus clearly precluding a fair comparison. In addition, the review included studies that compared exercise plus well-established treatments versus other well-established forms of treatment, such as pharmacological antidepressants (Blumenthal et al., 1999). As a result, these limitations directly affected the effect size (ES), producing a “shrinkage” effect on the efficacy of exercise for depressive symptoms when compared to previous meta-analyses (Ekkekakis, 2015). In addition, separate subgroup analyses of studies that assessed the effects of exercise on Beck depression inventory (BDI) (Beck et al., 1961) scores were also criticized regarding the inclusion criteria (e.g. including a trial which used the Hamilton HAM-D (Hamilton, 1967) scale for depression and not the BDI (Blumenthal and Doraiswamy, 2014, Cooney et al., 2014)).
No recent (within the last decade) comprehensive meta-regression analyses have been conducted investigating exercise and depression. Previous meta-analyses (Craft and Landers, 1998, Rethorst et al., 2009) evaluated the moderating role of sample characteristics, such as a diagnosis of major depressive disorder (MDD), which were found to be significant moderators of the antidepressant effects of exercise. However, a number of additional eligible studies have since been published.
Another limitation within the available literature investigating the effects of exercise on depression is that no previous meta-analyses have adjusted for publication bias, which is a considerable threat to the validity of any such synthesis (Ioannidis et al., 2014). Previous studies of psychotherapy for depression have demonstrated that publication bias is evident in RCTs, and effect sizes have consequently been overstated (Cuijpers et al., 2010). It remains unclear, however, if publication bias threatens the validity and interpretation of the exercise as a treatment for depression literature.
The present review sets out to address these limitations. Specific aims were: (1) to establish the updated effects of exercise on depression comparing exercise versus non-active control groups, (2) to identify moderators through meta-regression analyses, including sample characteristics (sex, use of medication and severity of baseline symptoms) and exercise intervention variables (length of the trial, frequency) that could impact the effects of exercise on depression, (3) to investigate, through subgroup and sensitivity analyses, the magnitude of the effects of exercise considering study quality, group format, setting, intensity, type, supervision, presence of clinical co-morbidities, type of publication and diagnosis of MDD, (4) to assess the influence of publication bias on the reported effects of exercise on depression, and (5) to quantify the strength of the existing evidence by calculating the number of negative studies required to nullify the pooled ES of the analyses performed.
Section snippets
Methods
This systematic review is in line with the PRISMA statement (Moher et al., 2009) and the MOOSE guidelines (Stroup et al., 2000).
Search results
In the first stage of the search strategy, 35 RCTs were identified from a previous review (Cooney et al., 2013). In the second stage, following the removal of duplicates, 819 potentially relevant articles were identified. At the full text review stage, we reviewed 76 articles (N = 35 from stage 1 and 41 from our stage 2 searches) and 45 were excluded with reasons (details summarized in Fig. 1). There were 30 full texts that met the eligibility criteria (Blumenthal et al., 2007, Brenes et al.,
Discussion
This meta-analysis found large antidepressant effects of exercise on depression when compared to non-active control conditions (e.g. studies that did not compare exercise versus alternative treatments). The anti-depressant effect of exercise was higher for studies that included participants diagnosed with MDD. Moreover, our adjusted analyses demonstrate that publication bias generally resulted in an underestimation of the positive effects of exercise. Larger effect sizes were found for
Acknowledgements
The authors would like to thanks to Coordenação de Aperfeiçoamento de Pessoal de Nível Superior.
References (73)
- et al.
Exercise or basic body awareness therapy as add-on treatment for major depression: a controlled study
J. Affect. Disord.
(2014) Honey, I shrunk the pooled SMD! Guide to critical appraisal of systematic reviews and meta-analyses using the Cochrane review on exercise for depression as example
Ment. Health Phys. Act
(2015)- et al.
Combined exercise and cognitive behavioral therapy improves outcomes in patients with heart failure
J. Psychosom. Res.
(2010) - et al.
Effects of aerobic exercise on sexual functioning in depressed adults
Ment. Health Phys. Act
(2009) - et al.
Publication and other reporting biases in cognitive sciences: detection, prevalence, and prevention
Trends Cognitive Sci.
(2014) - et al.
Effects of adjunctive exercise on physiological and psychological parameters in depression: a randomized pilot trial
J. Affect. Disord.
(2015) - et al.
Moderate exercise improves depression parameters in treatment-resistant patients with major depressive disorder
J. Psychiatr. Res.
(2011) The file drawer problem and tolerance for null results
Psychol. Bull.
(1979)- et al.
Exercise and severe depression: preliminary results of an add-on study
J. Affect Disord.
(2011) - et al.
Exercise and severe major depression: effect on symptom severity and quality of life at discharge in an inpatient cohort
J. Psychiatr. Res.
(2015)