The purpose of the review was to examine the efficacy of regular PA as a potential adjunctive treatment to reduce psychiatric symptom severity in individuals with SMI, defined as schizophrenia spectrum disorders, bipolar disorders, and major depressive disorder diagnosed based on DSM or ICD criteria. While there is an abundance of evidence on the benefits of PA on mood and wellness in the general population, and the physical health benefits of all people, there is a paucity of research on the efficacy of PA in reducing psychiatric symptom severity in individuals with severe mental illness. This is unique population that faces significant barriers to participation in regular PA yet could benefit immensely (9, 27). The physical health benefits of regular PA among this population are well established and were therefore not a focus of this review.
The sample included in the review was quite heterogenous, differing in gender, diagnosis, geographical location, psychiatric symptom measures, PA intervention, and including both inpatient and community-based populations. Direct comparison of groups and any sub-groups conclusions could not be made, but there is sufficient evidence to support that PA interventions generally result in small improvements on multiple indicators of psychiatric symptom severity. While effect sizes were small, many of the studies found participants had some improvement in psychiatric symptom severity and none of the studies reported any negative outcomes or harm resulting from the PA intervention. This lack of harm and potential for physical and psychiatric health benefit was consistent across an immensely diverse sample, providing some robust support and evidence for promotion of PA as an adjunctive treatment for individuals with SMI. Even though effect sizes are small, there is a clear beneficial effect of PA in this population and no harm. Further research should explore the specific PA interventions, frequency, and duration for maximal reduction of psychiatric symptom severity so that recommendations can be made on how to use PA as an adjuvant intervention for people with SMI.
Of the ten studies included in the final review, 40% of studies scored as weak and 50% scored as moderate on the quality rating assessment. The lower quality scores were typically due to selection bias and lack of blinding. Because the studies focus on individuals with severe mental illness, a vulnerable population, it is likely that ethics created some logistical barriers to study design. Some of the studies explicitly refer to informing participants of the research question at the request of the ethics board. Active recruitment may also have been denied by ethics committees, particularly in in-patient settings, which would limit researchers to more passive recruitment such as displaying recruitment posters. While this protects potential participants from real or perceived abuse of power, it does introduce selection bias and would make recruitment of larger numbers challenging, particularly among individuals with psychotic disorders or severe depression, where the motivation and organization needed to independently contact researchers may be impacted. In future, it is critical that research include vulnerable populations, such as persons living with mental illness, so that we can better understand and treat these conditions. The American Psychiatric Association has summarized the aforementioned challenges and provided a list of recommendations for future study designs to ethically include this population (28).
Several studies also had small sample sizes, which was also likely due to the pragmatic challenges of recruiting participants, conducting research with individuals with SMI, and the ability of those individuals to consent to and participate in a study for the duration of the study. As mentioned previously, low motivation, anhedonia and other common symptoms of the disorders being examined are likely to have created barriers in recruitment, participation, and retention. Future research could potentially circumvent some of these barriers by doing retrospective studies on physical activity initiatives in settings such as tertiary care where residents are required to partake in daily activity programs and selection bias is therefore less likely to influence results.
Another limitation of many of the studies was a lack of control groups. Only three of the ten studies included in the final analysis had control groups that received no intervention. Two of the studies, Strassnig et al. (2015) and Woodward et al. (2018), had no control group. Five of the studies had active control groups. For example, both Greer et al. (2016) and Streeter et al. (2017) administered the same intervention to the experimental and active control groups, but at different levels of intensity and frequency. They then assessed dose-dependent effects of the intervention. The efficacy of the intervention itself cannot be quantified in this type of study design because both groups will have gained some benefit, and the effect size will be diluted. Oertel-Knöchel et al. (2014) compared an exercise intervention group with an active control group enrolled in a relaxation program. This design makes it difficult to discern if improvements are due to PA or enrollment in a program that provides structure, support, social connection, or other benefits to participants. This study design was likely due to ethics requirements when working with vulnerable populations. Future research should consider how to better control confounding variables in the study design, perhaps by using waitlist control groups instead of active control groups, or by comparing group PA interventions to individual PA interventions.
Similarly, several studies compared the intervention group with an active control group who were engaged in a different activity. Malchow et al. (2015) compared an aerobic intervention to an active control group who partook in table soccer instead of the aerobic exercise. Su et al. (2016) compared an aerobic exercise intervention with a stretching/toning active control group. The Woodward (2018) study design was even more convoluted as it compared an aerobic intervention group with a weight-bearing exercise intervention group. Depressive symptomology improved in both groups, and there was no non-PA intervention control group. With these types of study designs, particularly because there is a paucity in knowledge about the effects of any type of PA on psychiatric symptom severity and the mechanisms underlying these effects, it becomes difficult to parse apart causation. It is possible that participants are benefiting simply by being enrolled in a program. The improvements in psychiatric symptom severity could be due to the structure, social connection, and mental stimulation of program involvement and not the PA intervention. We first need to understand the effects of PA on psychiatric symptoms in general, and then begin to parse apart the potentially unique effects of different types of PA interventions.
Several of the studies included interventions in addition to the PA intervention. Malchow et. Al (2015) and Oertel-Knöchel et al. (2014) both used PA interventions in conjunction with computer assisted cognitive remediation training using ‘COG-PACK’. It is possible then that improvement in symptoms is due, fully, or partially, to benefits of the cognitive training. The effects of the PA intervention cannot be clearly quantified or assessed. Kang et al. (2016) included a social skills and self-management training components to their Thai-Chi intervention that makes it unclear if improvements were due to PA or non-PA interventions.
Perhaps the most complex study was the Jakobsen et al. (2017) study, which compared the CHANGE intervention with “care coordination” and “treatment as usual” control groups. The CHANGE intervention was a 1-year affiliation with a coach who would take a multifaceted, holistic, and individualized approach to wellness that included encouraging daily PA, but also addressed dietary issues, smoking, and other lifestyle factors. The care coordination group provided participants with a specially trained psychiatric nurse to coordinate access to primary health care, and the “treatment as usual group” were given access to a general practitioner and annual metabolic screening. With the complexity of these interventions and no measurable, controlled, or consistent PA intervention, it is not possible to draw causative connections between PA in the CHANGE group and psychiatric symptom severity.
Additionally, there is the pragmatic challenge of being able to test PA efficacy in psychiatric symptom management without withholding treatment for these conditions, which would of course be unethical. For example, Siqueria et al. (2016) assessed the efficacy of an aerobic exercise intervention in decreasing depressive symptomology in individuals with major depressive disorder but did so by looking at the dosage of antidepressant medication required to manage symptoms. They found that the exercise group required lower doses of medication. This is clearly not a direct measure of the efficacy of the intervention in reducing symptomology, but it would be unethical to allow one group to experience more severe symptoms by withholding medication or keeping medication doses controlled and at subtherapeutic levels so that the PA intervention can be assessed. If medication were controlled at a therapeutic level, symptoms would be controlled, and it would not be possible to detect improvements from PA.
Despite this review compiling broad, global evidence unanimously supporting PA as an adjuvant treatment for SMI, there were some limitations. We searched the three main databases for this domain for all English-language articles, but articles not listed in one of these databases or published in another language would be not have been retrieved.