Impact of exercise training on symptoms of depression, physical activity level and social participation in people living with HIV/AIDS: a systematic review and meta-analysis

The following eligibility criteria were used to select the studies for the review:

An extensive search strategy to spot eligible studies was implemented in two phases in line with the PRISMA protocol, including; (i) the search of the bibliographic database and grey literature, and (ii) the choice of studies for inclusion based on eligibility criteria. Searches involved some combinations of search terms from medical subject headings (MeSH) and keywords with a mix of symbolic logic within the title, abstract and text for the population, intervention, control and outcomes, first in a preliminary or pilot search to determine the sensitivity of the search strategy. The search strategy was used differently for the three selected study outcomes. PubMed search strategy is shown in Additional file 1: Appendix I. This strategy was modified to the syntax and subject heading of other databases. Studies were searched in PubMed, Emcare, Cochrane Library, Embase, CINAHL, AMED, PsycINFO and MEDLINE. Additional searches were made of the reference list of identified studies. This procedure was implemented by the recommended guidelines of the Cochrane Handbook for Systematic Reviews [47], and advice for Health Care Review by the Centre for Reviews and Dissemination [48].

The search results were exported to the RefWorks™ manager where the Checks for duplication of the identified studies were done. Using the RefWorks™ manager, the bibliographic records were exported to Microsoft Excel 2007 for organization and sorting of articles according to specific eligibility criteria.

The screening was performed in two phases. V.U. (reviewer 1) carried out the first phase of screening based on the title and abstract to spot articles that met the eligibility criteria. The screening results were independently cross-checked by I.F.O (reviewer 2). The two reviewers then read through the full text of the chosen studies for further screening, using the eligibility criteria. In order to reduce assessor bias, differences in opinions about inclusion or exclusion of any identified study were resolved either through discussion and reflection or by consulting with D.I.S (reviewer 3). The reasons why studies were excluded were adequately stated, and a description of the study selection procedure is presented in Fig. 1.

The collected data were analysed using the review manager (RevMan) to provide an answer to the review question about the impact of physical exercise training on depression, PAL and social participation in PLWHA. Appropriate statistical techniques were applied for each of the study outcomes. The quantitative data synthesis was done by pooling the mean values of the outcomes, standard deviation, and the number of participants in the included studies. These were used to compute the post-intervention standardised mean difference (SMD) for the variables following the standard analysis procedure in the Cochrane meta-analyses. The SMD was determined where different outcome measuring tools/units were used in the included studies. Analysis of continuous variables was done using the weighted mean difference when outcomes were uniform, with a 95% confidence interval. The overall effects of the interventions were determined, evaluated and combined in a proof table. Interpretation of SMD was in accordance with a previous recommendation [50]: small = 0.00–0.39, moderate = 0.40–0.70, and large ≥ 0.70. Alpha was set at p < 0.05.

Characteristics of the retained studies were arranged by year of publication and presented in a tabular form providing information on authors’ references, sample size, age, setting, data collection format, outcomes, components of the intervention, component of the control, format and provider of the intervention, intervention and follow-up periods, and results. Three meta-analyses to find pooled effect sizes across studies were done, using a random-effects model. Cochrane’s χ² test (10% significance level) and Higgins I² were applied in assessing the heterogeneity of the data, and values of 25%, > 25–75%, and > 75% indicate low, medium and high heterogeneity, as specified in the Cochrane Handbook for Systemic Reviews of Interventions [48]. Examination and documentation of outcomes were made using the primary outcome. Studies that showed clinical or methodological resemblance in terms of design, intervention, and comparator(s) were pooled together for meta-analysis using a random-effects model. Heterogeneous studies were analysed and interpreted using narrative synthesis to determine the relationship and results between and within the studies included in this review. This was done following the recommendation of the Centre for Reviews and Dissemination [47]. A narrative synthesis was done by determining the relationship across the studies. The key concepts in the studies were compiled, matched and differentiated, expressing the studies into one another and adding-up the translations to identify the exact concepts that extend across individual accounts, and which was used to frame an understanding of the effects of physical exercise on depression, PAL and social participation. Sensitivity analysis was done after the primary quantitative synthesis to evaluate the effect that studies with a high risk of bias have on the general outcomes when they are introduced or removed as part of the analysis.

This review included five publications on depression [43, 51‐54], seven publications on social participation/functioning [54‐60] and two publications on PAL [61, 62]. All the studies included in this review were carried out in the post-HAART era. Altogether, a total of 779 participants was involved across the thirteen studies included in this review.

Reasons for the exclusion of eight studies following full-text screening included: a pilot study (n = 2), non-randomized control trials (n = 1), had missing data whose author(s) failed to provide the missing data on request (n = 1), questionnaire bias (n = 1), included seronegative patients (n = 1), used aerobic exercise training as control (n = 1) and did not study any of the outcome measures of interest (n = 1) (Fig. 1).

Table 1 presents the study characteristics of the thirteen publications included in this review, and further details are provided below:

This review included 13 studies with 779 participants, with a gender distribution of 378 males and 310 females, while 91 participants (11.68%) did not state their gender (Table 1). Only 596 (76.51%) participants completed the study while 183 (23.49%) withdrew. The male/female ratio is approximately 1.2:1. Participants in the studies were 18 years of age or older. Ten studies included participants receiving antiretroviral therapy [43, 51, 53, 54, 56‐60, 62]. Nevertheless, Neidig, Smith [53] and Jaggers, Hand [43] reported that only 75% of participants in their studies (35 out of 60) and 38.71% (36 out of 93), respectively, were using ART. However, three studies [52, 55, 61] did not report the ART status of the participants. Similarly, ten studies did not report the HIV staging of participants [43, 52, 54, 56‐62]. On the other hand, two studies [51, 53] reported that according to the Centre for Disease Control's classification of HIV infection, the participants' disease was in categories A and B, i.e., in the asymptomatic and symptomatic stages, respectively. Additionally, one study [55], reported that all the participants were in the non-AIDS category (i.e., “category A"). However, one study [43] reported that the participants' HIV staging ranged from "asymptomatic" (63%), symptomatic (10%), AIDS (25%), and undisclosed (22%).

Further details on the participants' characteristics in this review are provided below:

The risk of bias within the included studies is provided in Table 2. Performance bias (the lack of subject and therapist blinding) was the major source of bias in all the studies included. Based on the PEDro scale, ten studies (76.92%) were considered fair/moderate quality [43, 51‐53, 55, 57‐61]. Two studies (15.38%) were rated as high quality studies [56, 62] and one study (7.69%) was rated as low-quality [54]. Further details are provided below:

This review carried out three meta-analyses for the included studies on depression, PAL and social participation.

The combined evidence from the included studies seems to support the potential of physical exercise training to ameliorate the symptoms of depression in PLWHA. The synthesised evidence in this review revealed that aerobic exercises or combined (aerobic + resistance) exercises performed 2–3 times/week, at 40–60 min per session, and for between 6 and 24 weeks, effectively ameliorated the symptoms of depression in PLWHA. However, there were three times lesser females than males in the studies included in this review for depression. This might have some implications as a previous epidemiological study [63] in PLWHA revealed that symptoms of depression were more prevalent in women than men. The effect size in the meta-analyses was large, and therefore both males and females would have benefitted from the anti-depressant- effect of physical exercise training.

Additionally, as the upper limit of age in the individual studies was not above 65 years, it is expected that age will not contribute significantly to the distribution of depressive symptoms among the participants. For depression, the sensitivity analysis excluded one study [54] for the non-equivalence of the control and exercise groups at baseline, and which resolved the medium heterogeneity of the primary meta-analysis. This suggests that some flaws exist in the randomisation process of the included studies, and need to be addressed in future studies. As the excluded study [54] was conducted in a developing country (Brazil), it is reasonable to assume that one of the challenges in conducting RCTs in resource-limited settings may be related to the availability of the technology for implementing randomisation, which may vary from the developed clime.

The findings of this study agree with the evidence from other systematic reviews [1, 3, 64] that physical exercise training ameliorates the symptoms of depression in PLWHA. Besides, several systematic reviews [42, 65, 67] present evidence that aerobic and resistance exercise intervention at various intensities have both stand-alone and combined positive effects on various aspects of mental health or psychological outcomes, such as depression [40, 66‐68], hope, desire to continue living [65], and health-related quality of life [65] in PLWHA. O’Brien et al. [3] presented evidence that performing an aerobic exercise or a combination of aerobic and resistance exercises, three times per week, for at least five weeks, improved the symptoms of depression in adults with HIV.

A recent systematic review among PLWHA by Heissel et al. [64] also revealed that physical exercise has a large effect on depression (p = 0.02) compared to controls. However, the review included nine studies on both conventional and unconventional exercises like yoga, which is considered a mind–body practice that integrates various body poses, breathing techniques, and utilises deep meditation [69]. Invariably, the synthesised evidence by Heissel et al. [64] cannot be strictly attributed to the effects of physical exercises alone, unlike our study which included only studies that investigated conventional exercises alone. The presence of confounding factors may explain why the results from Heissel et al. have a higher statistical heterogeneity (I² = 94%; p = 0.02) and weakens our confidence in the estimate of effect. In contrast, our study has a lower heterogeneity and highlights the positive effect of exercise on symptoms of depression in a more homogenous sample of PLWHA. A lower heterogeneity increases our confidence in the estimate of effects of physical exercise training on symptoms of depression, and point to a significant role for physical exercise in the management of depression in PLWHA.

We further explored the literature for theoretical projections on the molecular and morphological basis for the antidepressant-like effect of exercise training on vital brain targets. The literature revealed that aerobic exercises could increase the growth or size (through neurogenesis) and function of the hippocampus thereby helping the brain to exert control actions that minimise oxidative stress, attenuate inflammation [70] and ameliorate depression [71]. In essence, exercise creates differing effects to the actions of the facilitators of depression such as diminished neurogenesis, which in depressed individuals indicates hippocampal volumetric decreases [72], and predisposition to experiencing a depression episode [72]. Running on a treadmill also helps alleviate depression by stimulating the wingless (Wnt)-frizzled (Fz) signalling pathways which is an upstream regulator/inhibitor of glycogen synthase kinase 3 (GSK 3) that is linked to depression [73]. A significant increase in GSK-3 expression impairs hippocampal neurogenesis and increases depression risk [74]. Exercise, especially resistance exercise, promotes the level of serotonin in the brain of humans, which facilitates cell proliferation and neuroplastic changes associated with antidepressant effects [75]. As well, exercise increases synaptic plasticity in hippocampus-prefrontal cortex neuronal pathways [71], which are typically compromised by depression [76]. Thus, in order to improve how depression is managed in PLWHA, the significance of exercise in promoting synaptic plasticity (when antidepressant treatments do not increase cell proliferation) and in maintaining brain function by stimulating the Wnt signalling channel, should be explored.

As far as we know, this is the first systematic review to demonstrate the effects of physical exercise on PAL in PLWHA patients. Though the meta-analysis from the two studies [61, 62] included for PAL showed no evidence that exercise training improves the PAL in PLWHA, however, the findings of both studies revealed a consistent trend of improved post-intervention PAL (though not significant) among the experimental groups compared to the controls. Since the overall effect was not significant, it implies that physical exercise training may not improve the PAL among PLWHA compared to no exercise group; aerobic exercise compared to normal routine activity group; aerobic and resistance exercise compared to other control groups. Nevertheless, there are only a few studies done in this area making it difficult to form a scientific opinion on this matter, especially as the two studies [61, 62] included in this review revealed a mixed result and therefore, lack of scientific consensus in this area.

McDermott et al. [61] reported no significant difference (p > 0.05) in the PAL between the experimental and control groups after 16 weeks of exercise intervention. However, the study did not show any deterioration in the PAL as well. This implies that exercise intervention should have a protective effect in maintaining the PAL even if it does not significantly boost it. However, the results of the study should be interpreted with caution, because the sample size was small (n = 13) and the number of participants that completed the study was even smaller (n = 11). In essence, the study was underpowered to detect the statistical difference between the two groups. Invariably, there is a possibility that the true effects of exercise training on PAL in PLWHA were underestimated in the study due to type II error. More so, there is no report regarding the effect size to estimate the clinical significance of their findings. On the other hand, Roos et al. [62] demonstrated that physical exercises significantly boosted the PAL of PLWHA within six months, but not beyond 6–12 months, whereby the compliance of the participants was lower than the first 6 months. Consequently, the findings demonstrate that unsupervised pedometer-based physical activity intervention can translate into tangible health benefits. This approach can be scaled-up to population-level to enable more active lifestyles for PLWHA if compliance is high as was the case from 0 to 6 months. Nonetheless, it is not certain to what extent an unsupervised physical activity behavioural intervention approach will translate to a sustainable behavioural/lifestyle change considering that the participants recorded a low compliance level between 6 and 12 months. It probably explains why the post-intervention assessment of PAL showed no significant difference between the intervention and control groups.

The above findings highlight the need to structure the physical activity review appointments, behavioural evaluations and goal-setting sessions within a common theoretical framework when implementing any walking strategy to tackle the barriers to physical activity [77]. Such a theoretical framework aids the individuals to find barriers to walking, and make choices on how to alter and improve their walking behaviour [78] using various strategies. McDermott et al. [61], and Roos et al. [62] did not apply this approach and seemed not to have implemented regular physical activity consultation sessions in their studies. This may partly explain why their studies were unable to find a significant difference in the PAL when the intervention group was compared to the control group. Notwithstanding this gap, the higher post-intervention PAL (though not significant) compared to the baseline value buttresses that physical exercise training boosts PAL in PLWHA. It reaffirms the evidence from McDermott et al. [61] that physical exercise training may have a protective effect in preventing a decline in the PAL of PLWHA even if it does not significantly increase the PAL. Importantly, the quality of the study by Roos et al. [62] is high and the sample size (n = 84) is sufficiently powered to detect the difference between the groups but was affected by the high attrition rate (> 15%), or else, the findings would have been more significant. Overall, the paucity of literature, apart from the observed flaws in the two studies highlighted above, warrants the conduct of more high-quality RCTs to determine the true estimate of the effect of exercises on PAL in PLWHA, which is important to guide practice.

The available evidence demonstrates that the overall effect of exercise on social participation in PLWHA is not significant. However, we found a high attrition rate (> 15%) in all the included studies except three [57‐59], which may influence the outcome of the findings. Besides, the heterogeneity of the primary meta-analysis was high, and likewise the sensitivity analysis. Therefore, the data from the included studies do not seem to represent the true estimate of effect in a homogenous population of PLWHA. The heterogeneity of the data may be related to the diverse measuring tools for social participation/functioning employed across the included studies. Also, there is a selection bias as five [54, 56‐59] of the seven papers did not implement the concealed allocation of the participants into the study groups. One of the studies with selection bias [58] reported a significant post-intervention decline in social participation in the intervention group than the control. However, the true estimate of effect is in doubt because the baseline value of social participation was significantly higher in the control group than in the intervention group probably due to flaws in the randomisation procedure. Despite this limitation, exercise training did not reduce social participation below its baseline value in the intervention group. In essence, social participation neither deteriorated nor significantly improved among the exercising participants compared to control. Also, there is a likelihood that age might have some effects considering that older adults, up to 86 years of age, were involved in the included studies. Hence, the findings of our study do not agree with the results of previous studies in non-HIV infected population including a social experiment [79], a systematic review [80], a descriptive cross-sectional study of individuals with disabilities [82], and an RCT of stroke survivours [81]. Similarly, our study presented evidence at variance with the estimate of effect reported by another systematic review by Vancampfort et al. [83] which included only studies involving PLWHA but was not restricted to RCTs.

All the cited studies above [79‐83] identified a link between physical activity profile and social participation probably due to the effects of physical activity on mental health, especially mood [84‐86]. However, several limitations related to study design, and the involvement of mostly non-HIV infected population do not allow for the comparisons of the results between our study and the above-cited studies. Nevertheless, the importance of our study is supported by a previous view that HIV and rehabilitation research should focus on social participation [87]. In an HIV population, it is especially important as social stigmatization is problematic for PLWHA in different countries of the world [88]. Social stigmatisation in PLWHA may lead to low self-esteem, mood disorders, and withdrawal from social activities or social isolation [88] resulting in a sedentary lifestyle or a decline in ambulatory function. Meanwhile, the ambulatory function is often restricted with polypharmacy, and a subsequent rise in complications associated with HIV infection [86] may add to the restrictions or lack of participation in physical (sedentary lifestyle) and social activities. The resulting multi-system (neurological, musculoskeletal, cardiopulmonary and metabolic) dysfunction may impair walking function, as well as compromise the quality of life [26]. Limited social participation may also be related to a feeling of low self-esteem or rejection [89]. Therefore, any intervention that may improve mental health by improving mood, and self-esteem in PLWHA, may also improve their physical functioning and likewise social participation. Knapen et al. [90] analysed four meta-analyses on the effects of physical exercise on mental and physical health in people with mental health problems. They found that exercise training resulted in a better quality of life, reduced emotional stress, enhanced self-esteem, and body image which have implications for a social relationship, and may have translational benefit for PLWHA. It emphasizes the point earlier made by Dianatinasab et al. [91] that “keeping up physical activities and exercises are key strategies in medical and social care of PLWHA.”

There are several limitations to this study. First, the variables of interest are constructs, which are somewhat subjective and therefore difficult to measure objectively or accurately. Besides, the self-reported psychometric instruments used in data collection for the included studies are prone to recall bias. Importantly, the small sample size in some of the included studies will likely introduce bias for type II error, and likewise, the small number of studies that were included in this review that met the eligibility criteria. The application of the standard mean difference provided a comparative basis to evaluate the effects of physical exercise training on the outcomes across the studies, where diverse measuring tools were used. However, there are difficulties applicable to the interpretation of the SMDs, which may limit the true estimation of the effects of the intervention across the studies. For instance, the interpretation of the SMDs varies across instruments depending on whether improvement in a given instrument is associated with lower or higher scores on the outcome measure [53]. This is quite tricky when the SMDs have overlapping confidence intervals, which increases the failure to reject the null hypothesis. This also means an increase in type II error than the corresponding hypothesis test. Consequently, there is a need for more robust, rigorous and high-quality studies that will measure each variable of interest using more than one tool to enable realistic comparison of data across studies.

Evidence regarding the impact of exercise training on depression favours exercise when compared to control and suggests that the symptoms of depression are ameliorated by exercise training. These findings have important implications for practice, especially in sub-Saharan African where despite the existing evidence of the beneficial value of physical exercises/activity in improving mental health, yet physical exercises/activity is neglected as a routine rehabilitation modality in the mental healthcare systems [83]. Importantly, mood disorders, which is common in PLWHA, have also been linked to poor adherence to medications and consequently, poor health outcomes [1, 2, 36‐39]. If physical activity/exercises ameliorate the symptoms of depression, it may also improve adherence to medication and likewise other related health outcomes. Therefore, public health policies and initiatives designed to increase participation in exercises may have the potential to improve mental health and general wellbeing among PLWHA.

Across the included studies, there were notable potential sources of bias related to baseline comparability, incomplete outcome reporting, small sample size, short intervention duration, non-implementation of blinding of therapists and assessors. Also, diverse outcome measuring tools with different validity and reliability values were used in the included studies, which made data comparison across the studies more difficult. Some studies that evaluated the effects of exercises on depression and PAL were underpowered to detect the differences in the mean between the intervention and control groups. Some of the studies did not provide information on the adherence rate and effect size to enable the estimation of the effect of the physical exercise training on the variables of interest in the PLWHA. Similarly, information on whether the participants were on antidepressant medications before the study was not provided in any of the studies. This makes it difficult to determine whether the antidepressant-like effects of physical exercises were independent or amplified in those taking antidepressant medications. These weaknesses may affect the validity of the findings of the various studies included in this review. Meanwhile, the paucity of RCTs on the impact of physical exercise training on PAL made it difficult to form a scientific opinion on its’ effect.

The variability in the number of participants across the studies was high for social participation, ranging from 21 participants in one study [59] to 100 participants in another study [57]. The same applied to studies on physical activity level which included participants ranging from 11 participants in one study [61] to 84 participants in another study [62]. In contrast, the studies on depression had the least variability of participants, ranging from 40 participants in one study [52] to 93 participants in another study [43]. This may contribute to the heterogeneity of the data across the studies, and might explain the substantial heterogeneity in the studies related to social participation and PAL. Additionally, the substantial heterogeneity suggests that there are some clinical or methodological differences, or both, across the studies included in the review. The is buttressed by the fact that the included studies showed variances in the study characteristics such as gender, location of study (developed versus developing countries), the type and dose of physical exercise training, measuring tools, sample size as well as non-equivalence of the study groups at baseline, and high attrition rates. Therefore, the variances also add to the clinical or methodological differences and suggest some flaws in the trials design, documentation and implementation which should be addressed in future studies.