Depression and Objectively Measured Physical Activity: A Systematic Review and Meta-Analysis

Depression is a major contributor to the overall global burden of disease, with high prevalence and relapse rate. Several factors have been considered in order to reduce the depression burden. Among them, physical activity (PA) showed a potential protective role. However, evidence is contrasting probably because of the differences in PA measurement. The aim of this systematic review with meta-analysis is to assess the association between objectively measured PA and incident and prevalent depression. The systematic review was conducted according to methods recommended by the Cochrane Collaboration and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Relevant papers published through 31 August 2019 were identified searching through the electronic databases PubMed/MEDLINE, Excerpta Medica dataBASE (Embase), PsycINFO, Scopus, Web of Science (WoS), and the Cochrane Library. All analyses were conducted using ProMeta3. Finally, 42 studies met inclusion criteria. The overall Effect size (ES) of depression for the highest vs. the lowest level of PA was −1.16 [(95% CI = −1.41; −0.91), p-value < 0.001] based on 37,408 participants. The results of the meta-analysis showed a potential protective effect of PA on prevalent and incident depression.


Introduction
Depression is one of the major leading causes of disability worldwide, affecting approximately 400 million people [1], with 9% of men and 17% of women experiencing depressive symptoms at least once in their life. Mainly due to social prejudices, depression continues to be frequently under-diagnosed and inadequately treated [2]. Depression can have several negative consequences, being characterized by sad mood and/or loss of interest, affecting thoughts, feelings, behaviors,

Materials and Methods
We conducted this systematic review according to the methods recommended by the Cochrane Collaboration [16] and to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines [17] and documented the process and results in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [18]. The review protocol has been registered on PROSPERO [19], the International Prospective Register of Systematic Reviews funded by the National Institute of Health Research (https://www.crd.york.ac.uk/prospero/).

Information Sources and Search Strategy
Studies were identified searching through the electronic databases PubMed/MEDLINE, Embase, Scopus, Web of Science (WoS), PsycINFO and the Cochrane Library. We combined the search strategy of free text terms and exploded MESH headings for the topics of depression, physical activity, objective measurements, and type of study. The strategy was firstly developed in PubMed/MEDLINE and then adapted for use in the other databases (Supplementary Table S1). Studies conducted on human subjects and published in English through 31 August 2019 were included.

Inclusion and Exclusion Criteria
We considered studies that investigated the relation between physical activity objectively measured and depression, both as a continuous and as a binary variable. Adult participants of both sexes were considered. As done before [20,21], both population-based and hospital-based studies were included. Among hospital-based studies, inpatients, day-hospital, and outpatient subjects were included while emergency care records were excluded as considered non-representative. All experimental and observational study designs were included apart from case reports. Narrative and systematic reviews, letters to the editor and book chapters were excluded. Table 1 shows a detailed description of inclusion/exclusion criteria according to the Population, Exposure, Outcomes and Study design

Study Selection and Data Extraction
Identified studies were independently reviewed for eligibility by two couples of authors (VG, LB, MM, SC) in a two-step process: a first screening was performed based on title and abstract, while full texts were retrieved for the second screening. At both stages disagreements by reviewers were resolved by consensus. Data were independently extracted by three authors (LB, MM, SC) and supervised by a senior author (VG) using an ad-hoc developed data extraction spreadsheet. The data extraction spreadsheet was piloted on 10 randomly selected papers and modified accordingly. As done before [23][24][25], both qualitative and quantitative data was extracted from the original studies. Qualitative data recorded included the following items: name of first author and year of publication, country where the study was conducted and period during which the study was performed, device used to measure PA and tool used for depression diagnosis. Moreover, characteristics of the subjects were recorded (e.g., age, gender, comorbidities). Quantitative data extracted includes: sample size, number of participants lost (attrition), duration of PA measurement, distribution of depressed participants in the sample, level of PA performed and the results estimating the association between PA objectively measured and depression.

Quality Evaluation
The quality evaluation of the included publications were independently assessed by two authors using the New-Ottawa Scale [26] for observational studies and Cochrane Collaboration tool for trials [27].

Meta-Analysis
We pooled individual studies data using ProMeta3 ® (Internovi, Milano, Italy) software. Due to heterogeneity, a random effects meta-analysis was employed. In order to reduce the heterogeneity, two sensitivity analyses were conducted, considering the following items: (i) study design, (ii) participants' comorbidities. Moreover, a subgroup analysis by gender was conducted in order to estimate potential different effects among the two groups. We assessed publication bias with the visual inspection of a funnel plot [27] and the Begg [28] and Egger [29] tests.

Literature Search
A total of 4279 articles were retrieved. After a preliminary screening 670 articles were excluded because of duplicates, 409 not original papers (reviews, letters to the editor, editorials, protocols, etc.), and 2796 covering a different topic. After title and abstract screening, a total of 192 full-text articles were consulted, while at the end of the screening process only 41 were included in the systematic review . As it was not possible to extrapolate data from one study, it was not included in the quantitative evaluation [67]. Figure 1 shows the selection process. Two studies reported separate data for men and women [49,54] and for this reason they were considered separately, resulting in 42 datasets being included in the meta-analysis. while at the end of the screening process only 41 were included in the systematic review . As it was not possible to extrapolate data from one study, it was not included in the quantitative evaluation [67]. Figure 1 shows the selection process. Two studies reported separate data for men and women [49,54] and for this reason they were considered separately, resulting in 42 datasets being included in the meta-analysis. The characteristics of the included studies are reported in Table 2. The majority of the studies were conducted in Europe (n = 18, 43%) and North America (n = 12, 29%). The first study assessing objectively measure PA and depression was published in 2004 [68]. The smallest sample size included in a study was of 23 participants [70], whereas the largest sample size was of 16,415 participants [62]. Twenty-six of the 42 datasets were cross-sectional (62%), eight trials (19%), six cohort studies (14%), and one case-control study (2%). The quality assessment of trials is reported in Supplementary Table S2. Thirty-two datasets (76%) used an accelerometer as the measurement device, while nine datasets (21%) used a pedometer. In almost all studies participants were asked to wear the device for 7 days, and even in cohort studies PA was measured only at baseline. With regard to depression, heterogeneous tools were used to make diagnosis, such as the Hospital Anxiety and Depression Scales (HADS), the Patient Health Questionnaire-9 (PHQ-9), the Beck Depression Inventory (BDI-II) and the Center for Epidemiologic Studies Depression Scale (CESD). Most of the time HADS was used (n = 11), followed by PHQ-9 questionnaire (n = 9); however almost all studies used a validated tool. At the same time, the results were expressed using different measures, as for instance Odd Ratio (OR), Relative Risk (RR), β coefficient (β) and Spearman's Rho (r). The characteristics of the included studies are reported in Table 2. The majority of the studies were conducted in Europe (n = 18, 43%) and North America (n = 12, 29%). The first study assessing objectively measure PA and depression was published in 2004 [68]. The smallest sample size included in a study was of 23 participants [70], whereas the largest sample size was of 16,415 participants [62]. Twenty-six of the 42 datasets were cross-sectional (62%), eight trials (19%), six cohort studies (14%), and one case-control study (2%). The quality assessment of trials is reported in Supplementary Table S2. Thirty-two datasets (76%) used an accelerometer as the measurement device, while nine datasets (21%) used a pedometer. In almost all studies participants were asked to wear the device for 7 days, and even in cohort studies PA was measured only at baseline. With regard to depression, heterogeneous tools were used to make diagnosis, such as the Hospital Anxiety and Depression Scales (HADS), the Patient Health Questionnaire-9 (PHQ-9), the Beck Depression Inventory (BDI-II) and the Center for Epidemiologic Studies Depression Scale (CESD). Most of the time HADS was used (n = 11), followed by PHQ-9 questionnaire (n = 9); however almost all studies used a validated tool. At the same time, the results were expressed using different measures, as for instance Odd Ratio (OR), Relative Risk (RR), β coefficient (β) and Spearman's Rho (r).

Sensitivity Analysis by Participants' Comorbidities
The sub-group analysis considering only the general population (without diseases), included 21 datasets, and the pooled ES was −1.

Discussion
The current systematic review with meta-analysis-which included 43 studies in qualitative evaluation and 42 studies in the quantitative analysis-provided data on the association between objectively measured PA and the risk of depression. Since some studies expressed data separated for gender, a total of 42 datasets have been considered. With the purpose of deeply understanding the strength of the association between objectively measured PA and depression, a sub-group analysis by participants' comorbidity has been conducted. When studies assessing the association among participants with comorbidities were considered, the ES were not statistically significant (apart for COPD participants). However, prescription of adapted PA among participants affected by co-morbidities should be considered [71]. To the contrary, when only studies with general population (otherwise healthy people) were considered, the pooled ES was statistically significant, indicating an inverse association between PA objectively measured and depression (more PA was associated with lower risk of depression). A subgroup analysis by gender was conducted as well, showing a protective effect of PA only for women. However, this result should be considered carefully, since only three studies assessed PA and depression only in men, reducing the sample size.
These results are extremely important considering that depression is one of the leading causes of disabilities worldwide [1]. In the last fifty years a great concern was casted on physical health of depressed individuals. This could be due because physical exercise seems to improve several biomarkers implicated in depression (e.g., impaired neuroplasticity, autonomic and immune imbalances) [9]. In in-vivo models, physical activity showed a serotoninergic effect as some antidepressant medications [8]. Moreover, PA has demonstrated an effect on inflammatory processes, through the hypothalamic-pituitary-adrenal axis regulation involved in the development of depression [9]. Additionally, higher levels of brain derived neurotrophic factor have been found after physical exercise [10]. Lastly, the level of PA directly affects the upper limit of oxygen uptake which depends on the capacity of the cardiorespiratory system to transport oxygen to the organs, including the brain. A lower oxygenation of the brain may result in a chronic cerebral ischemia and, if the affected areas are involved in a mood regulation, this may increase the risk of depression [12].
In the last decades, several studies have shown that a healthy lifestyle, in particular the intensity and length of physical activity [72,73], are important in the prevention and treatment of depression [7]. In our analysis we could not assess the relation between severity of depression and intensity of PA, as in most of the primary studies included, severity of depression was not reported and PA intensity was expressed using different methods. The results from our review confirm the beneficial effect of PA on depression, especially for participants without comorbidities. In this regard, health education campaigns aimed to promote PA should be fostered [74][75][76], especially because approximately 40% of the adult population worldwide is insufficiently physical active [77]. However, in order to better interpret our results, another important aspect should be considered: indeed, even if several sub-group analyses have been conducted, the value of heterogeneity remained stably high. Although a sensitivity analysis including only datasets with otherwise healthy people has been conducted, the I 2 remained extremely high. However, a I 2 value higher than 90% means that heterogeneity is directly due to heterogeneity among studies, instead of sampling error [78]. Moreover, primary papers expressed the level of PA using different types of unit of measures and also the results were reported using different modalities. Even if the pooled ES has been estimated by log OR, allowing comparability, this underlying heterogeneity might have affected the assessment of the I 2 [79]. Another potential explanation of heterogeneity could be the different type of duration of measurement, the device used and the questionnaire adopted to diagnose depression. Furthermore, a variety of confounding variables were selected in original studies and, in order to control the results, we pooled the models with the highest level of adjustment.

Limitantions and Strengths
The main limitation of this systematic review is the high I 2 value that might reduce the generalizability of our results. Most studies are observational and based on cross-sectional analysis. Nevertheless, we performed sensitivity analyses only including trials and longitudinal studies, increasing the robustness of our results. Due to the high heterogeneity in reporting the level of PA performed by participants in original studies, it was not possible to identify a recommended level of PA. The inability to estimate an association between severity of depression and PA is another important limitation. The main strengths of this review are being systematic in nature and its comprehensive way to include the entire scientific evidence published so far on the main medical-scientific databases. Furthermore, the pooled ES was significantly large, based on 37,408 participants, and sub-group analyses have been conducted based on participants' comorbidity and study design. In the primary studies, diagnosis of depression was consistently based on the DSM criteria and was established by trained investigators using validated assessment scales mainly with interrater reliability.

Conclusions
To conclude, the results of this systematic review and meta-analysis clearly show a statistically significant protective effect of objectively measured PA on prevalent and incident depression. An increased PA is associated with lower risk of depression. The advantages of our study are several. Firstly, this study offers a systematic overview of previous studies assessing objectively measured PA and depression. Secondly, this study highlights the usefulness of objectively measured PA compared to self-reported one. Objectively measured PA is not only more precise in estimating duration, total amount, and intensity of PA, but indirectly it can also better strengths the association with some diseases, as depression. Thirdly, this study shows the importance to promote physical activity forasmuch it can help to reduce the high burden of depression in our society. Lastly, our findings are relevant for both policy makers and clinicians as physical activity is one of the cheapest, non-pharmacological treatment that might be prescribed to the general population with potentially major public health impact. Physical activity is important across ages and should be integrated into daily life.