Multidimensional analyses of the effect of exercise on women with depression

Abstract Background: The proportion of women is higher than men in depression. This is mainly due to women's physiological regulation is different from men, especially in puberty, menstruation, pregnancy, menopause, among others. Therefore, treating depressive women is still a health challenge. Besides, recent studies of exercise therapy have a more outstanding performance in treating depression, especially in contrast to drug therapy and psychotherapy. Its main advantages are convenience, quickness, no side effects, real-time, and long-term effectiveness. Objective: The aim of this study was to systematically review the clinical efficacy of exercise on women with depressive symptoms. Methods: Searching PubMed, The Cochrane Library, and Embase databases to collect randomized controlled trials about exercise in the treatment of depressive women. After literature screening, data extraction, and literature quality evaluation, the meta-analysis of acquirement data was performed with RevMan5.3 software. Results: A total of 2294 patients were included in 25 different articles totally. Meta-analysis shows that compared with the control group, exercise could relieve female depression (standard mean difference [95% confidence interval, CI] = −0.64 [−0.89 to −0.39], Z = 4.99, P < .001). Subgroup analysis shows that different types of exercise have significant effects in improving depression symptoms. Exercise therapy has better effect on depressive patients induced by physiology or disease than ordinary depressive patients. Conclusion: Exercise can significantly improve depressive symptoms in women.


Introduction
Depression is a common disease that manifests as a mental disorder. According to World Health Organization statistics, it is expected that depression will be the top leading cause of the global burden of disease in 2030. [1] Nearly 800,000 people die from depression every year. [2] During the onset of depression, the patient showed low mood, loss of interest, and lack of energy. At the same time, it is accompanied by anxiety, sleep disturbance, loss of appetite, self-guilt, inattention, and the depressed symptom that is not a direct effect of a medical condition. Not only psychological, but also social and physiological factors affect depression in different aspects. People with tremendous stress and dysfunction in life are more likely to suffer from depression. However, a series of diseases such as cardiovascular disease, diabetes, and cancer are also important causes of depression.
At present, the clinical treatments of depression are still focused on psychology and antidepressants. However, neither psychotherapy nor medication can meet the patient's urgent need for recovery. Other factors may interfere with the results during the long treatment period. [3] Therefore, more and more researchers focus on non-drug treatments and realize that changes in living behavior can help depressive patients recover. Numbers of studies have shown that exercise has a significant impact on depression, but it differs depending on exercise types. A metaanalysis [4] showed that exercise as a treatment has a moderate to significant effect on depressive patients. Moreover, some researches [5][6][7][8][9][10] show that yoga, Qigong, and Tai Chi can improve the symptoms of depression, but the effects of above methods are related to the coach's training level and exercise duration. In most cases, exercise as adjuvant therapy and other interventions (dietary regulation, drug therapy, among others) have a more remarkable effect. However, this is just a comparison between exercise as monotherapy and conventional care. When comparing to psychotherapy or medication, there is almost no difference in effect.
The prevalence of depression in women is higher than that in men, probably related to women's unique menstrual cycle and hormone secretion. [11,12] For women, depression often occurs around the regular onset of the luteal phase in the menstrual cycle. The inducing factors for women with depression are mainly changes in metabolism, sex, lifestyle, ovarian activity during adolescence, pregnancy, and menopause, accompanied by complex mood disorders. [13][14][15] Besides, women are susceptible to breast cancer and ovarian syndrome, which can also cause depression and anxiety. [16] The direction of previous meta-analysis researches usually focuses on the classification of exercise. The study by Kvam et al [4] analyzed the effects of exercise and non-intervention, psychotherapy, and medication by randomized controlled trials (RCTs) of different interventions. Schuch et al [17] selected different exercise types, such as moderate-intensity, aerobic, and high-intensity exercise. And part of the meta-analysis has subdivided the research objects, based on age, prenatal, postnatal, breast cancer, menopause, and so on. The metaanalysis of Gong et al [18] compared yoga as an intervention treatment with a control group (prenatal care, prenatal exercise, social support, among others), and the results support the efficacy of yoga on depression in pregnant women. Although Nakamura et al [19] partially confirmed the improvement of physical activity on depression in prenatal and postpartum women, the analysis results observed unexplainable heterogeneity. This may be related to the history of depression or the subject's mental activity. However, there is no summary of exercise for women with special conditions (prenatal, postpartum, menopause, breast cancer, ovarian syndrome, and so on) and depressive women without exceptional circumstances in the database.
Considering both the increasing influence of depression and the lack of knowledge in the database, it is necessary to update the content about the effect of exercises on depression in women. The present meta-analysis objective is to investigate the effect of exercise as a treatment for depression in women. We will examine the positive effect of different exercise and the effects of different types of depression in women (prenatal, postpartum, menopause, breast cancer, among others). We will also assess the effect of exercise on depression in different types of women.

Search
A search of PubMed, The Cochrane Library, and Embass database for RCTs on exercise on depression in women published from 2005 to the present were conducted on October 15, 2020. We also searched to retrieve all potential relevant unpublished reported materials and conference proceedings referred to the topic. We set the search time limit of nearly 15 years and limit the diagnostic criteria to control heterogeneity. Search terms include Depression, Depressive disorder, Exercise, Physical activity, Women, Female, Randomly, Randomized controlled trial, controlled clinical trial, and all types of exercise (eg, walking, running, aerobic exercise, resistance exercise, strength exercise, yoga, Qigong, Tai Chi).

Study selection
We selected all RCTs in which the language was English, and the study was exercising intervention on depression in women. There are 2 standards for participants: All adults aged >18 years with a diagnosis of depression as defined by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), International Statistical Classification of Diseases and Related Health Problems 10 th Revision (ICD-10) criteria or a confirmed depression scale; the participants limited to women. We exclude studies with men or a mixture of men and women and include articles of depression by diseases and physiology. Because depression often occurs in adults, our age limit is ≥18 years, and the ages of patients included in the exhaustive articles are 18 to 64 years. Besides, we include at least 2 groups of the articles. The experimental group was given exercise intervention, and the control group was given routine intervention measures like daily activities or lower-intensity exercise. We include documents that meet the above criteria and rule out those that meet the following criteria: intervention subjects are non-exercise conditions, and not only women; non-English articles; no experimental data can be calculated; experimental group is a combined intervention study; repeated research.

Outcome measures
We selected the following 4 related depression scale surveys: The Center for Epidemiological Studies Depression Scale (CES-D), [20] Beck Depression Inventory (BDI), The Edinburgh postnatal depression scale (EPDS), and The Hospital Anxiety, and Depression Scale (HADS). These are widely used on a mainstream scale, with a wide range of applications.

Literature retrieval
Two researchers conducted 3 rounds of screening participants obtained from the preliminary search in an independent doubleblind manner. The study did not meet the inclusion criteria or did not meet the quality requirements was excluded. Data extraction is performed using a unified feature table. The primary information includes the name of the first author, time, region, number of subjects, course of the disease, age, the primary intervention measure, evaluation indicators, and follow-up time. The 2 researchers then checked the results with each other. If there is a disagreement, the third researcher will arbitrate.

Effect estimators' level of evidence
We use the bias risk assessment tool recommended by the Cochrane Collaboration to evaluate the research bias risk. There are 7 items: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases. These 7 items are mainly evaluated from 5 aspects: the selection and implementation of the experimental plan, the measurement of the result data, the completion of the subjects, and whether the result report is objective and complete.

Data analysis
We use RevMan5.3 software for meta-analysis. Because the measurement data are continuous variables, the mean difference is used as the effect indicator (confidence interval is 95%). There is no significant difference in the baseline, and the endpoint after the intervention (M ± SD) is used as the main effect parameter. When we use the Q test, if P > .10, it shows no significant heterogeneity; in the I 2 test, if I 2 < 50%, it shows no significant heterogeneity. When there is no significant heterogeneity among the studies, the results should be analyzed using the fixed-effects model; for the studies with significant heterogeneity, the results should be analyzed using the random-effects model. Without affecting the authenticity of the analysis results, we select subgroup analysis according to factor classification to explore the source of its heterogeneity.

Additional information
With regard to ethical reviews, meta-analysis is based on data from published studies that do not involve patients and therefore do not require.

Study selection
We collected 1560 articles through preliminary search and then eliminated duplicate publications. Determining the research content by reading the titles and abstracts, we screened out 284 documents. After reading the complete text, we screened 98 articles based on their experimental design and output indicators. According to the inclusion criteria and quality evaluation, 25 articles were selected for qualitative analysis, and all of which  were finally determined, with 2294 participants. We conducted a meta-analysis on these documents. The database search was completed on October 2020. Figure 1 displays the research procedure and the flow of the study selection and inclusion.

Study characteristics and individual studies' results
Twenty-five studies were included in the qualitative and quantitative sustainability analysis. Their characteristics and main results are displayed in Table 1. For each of the studies  Table 1 Information extraction from Articles. Aibar et al, 2019 [21] Outpatients recruited through e-mail and telephone after contacting 2 associations of post menopausal women

Assessment of publication bias
Twenty-five studies mention the use of random allocation schemes, 13 articles detail the hiding methods of specific allocation schemes, 9 articles specify the implementation process of blinding, and 5 article illustrates the implementation of blinding results evaluation. All the 25 articles ensured that the experimental results were all omitted, and there was no selective reporting phenomenon. Both the study quality and risk of bias ratings are displayed in Figure 2 and

3.4.2.
Risk of bias across studies. The risk of bias across studies is by means of a funnel plot (Fig. 5). It reveals a low risk of publication bias.
3.5. Subgroup analysis 3.5.1. Exercise versus control. We performed 2 subgroup analyses to compare studies according to the exercise mode and the patient's condition.

Discussion
The main purpose of our study is to analyze the efficacy of exercise therapy for depression in women. The study showed exercise treatment, as a significant monotherapy, can effectively reduce the symptoms in the treatment of women with depression. Mood states rely on endorphin secretion. Exercise augments endorphin secretion which, in turn, reduces depression levels. [47] Also exercise may alleviate depression through neuromolecular mechanisms. [48] For women, depression often occurs during the active ovarian period. Adolescence, menstruation, pregnancy, and menopause are all "window of vulnerability" for depression, [49] but they may resist this period from a psychological level. As a long-term treatment plan for patients to follow, on the one hand, it effectively alleviates depressive symptoms to patients living discomfort. However, it can prevent psychological stress due to examination. Accordingly, compared with pharmacotherapy or psychotherapy of depression, exercise has similar beneficial effects.
Nevertheless, our analysis 1 shows the exercise has a significant effect on reducing depression in women (P < .001), but there is still high heterogeneity (I 2 = 85%). We guess that there are 2 reasons for this situation: different types of exercise will have different degrees of effect on depressive patients; the state of the patients may affect the therapeutic effect. We immediately subgroup analysis found that different types of exercise (aerobic exercise, resistance exercise, yoga, among others) have significant effects on depression in women, which is consistent with Kvam et al, [4] Cramer et al. [50] However, it is significant for the treatment of depressive patients with physiological conditions or other diseases (breast cancer, multiple sclerosis, primary Sjogren syndrome, and so on). It may be because exercise slightly improves the quality of life or reduce stress for ordinary depressive patients. But exercise can reduce the pain (dysmenorrhea, chemotherapy, among others) caused by physiological or              Figure 11. Forest plot of meta-analysis on the effect size of exercise on The Hospital Anxiety and Depression Scale on depression in women. CI = confidence interval, SD = standard deviation.  In addition, we will classify the different scales included in the articles and analyze them separately to examine the practicality of exercise therapy in more detail. We found that CED-S, HADS, BDI, EPDS statistics show that the treatment effect of exercise in depressive women is more prominent, and EPDS, BDI results are highly heterogeneous. We guess that it is related to the scales' own evaluation criteria and experimental conditions (exercise duration or research subjects).

Study or Subgroup
We conducted a subgroup analysis of all the articles that included EPDS, and the results show that the degree of improvement in depressive patients is related to the exercise duration. However, due to the limited number of included articles, the findings remain high heterogeneity. We suspect that the heterogeneity and the personal recovery progress of postpartum women about the reasons, the latter may also need evidence. EPDS applies only to postpartum women. The degree of    pain during childbirth itself will inevitably affect the posttreatment. But either because of the subjects or intervention itself or scale limitations, the movement of the efficacy of depression in women is still relatively significant. Finally, we conduct a subgroup analysis of the articles whose indicator is BDI. The results show that although the exercise has some effect, there are specific differences in the symptoms of patients with primary depressive mood and with disease complicated by depressed mood. It is similar to the results of a meta-analysis for breast cancer patients. [51] Although the final result is that exercise has a slight improvement in the depression of breast cancer patients, the number of references is limited, and the specific effects need to be studied in depth. From the intrinsic quality of the scale, BDI has high sensitivity. It is the final result that may fluctuate as a wide range of data appears. Wang et al [51] also conducted the meta-analysis on the application of BDI and found that BDI has a certain sensitivity. The questions it raised is too transparent for self-tested people, and it is more suitable for some patients who voluntarily participate or are familiar with their own conditions. Interestingly, we additionally found that some articles selected PSS to perform stress tests on patients. We also carried out a PSS analysis that showed a significant improvement in exercise for the role of patient pressure. Life stress is one of the risk factors for depression. Nowadays, many people choose to exercise as a decompression project in their free time.

Strengths and limitations
Our study included not just ordinary depression in women, but also through its contrast with depression by the disease or physiological problems (pregnancy, postpartum, menopause) caused analyzed and discussed, and provided new ideas for the treatment of depressive by exercise. We also analyzed the therapeutic effect of different types of exercise, and evaluated the effects of different scales on the treatment of depressive patients. At the same time, we also recognize several limitations. If we take all the experiment blindly for depression, it may cause psychological resistance, making it difficult to conduct and affecting the results. The assessment of depression scales is generally an indicator of subjective impressions. We also found that for the RCTs of depressive patients, although randomized, it is difficult to ensure that baseline indicators of each group are similar. This situation will cause high heterogeneity, making it difficult for us to judge the difference in efficacy by comparing endpoints.

Conclusions
The existing meta-analysis usually shows that the exercise intervention group reflects a more significant treatment effect than the control group. Therefore, exercise is an effective means to treat the women with depression, regardless of the cause including puberty, pregnancy, postpartum, menopause, or disease. However, due to the lack of research and publication of the highest quality deviation method, it is difficult to explain the findings. For now, the actual effect of exercise therapy in female patients with clinical depression requires further study in the future of more high-quality articles.