Depression is a major cause of physical and psychosocial illness and all-cause mortality 1. Depression affects more than 340 million people worldwide 2, with women at particular risk 3. Conventional management of depression has typically involved treatment from clinical professionals such as general practitioners, psychiatrists, psychologists, and has included counselling and anti-depressant medication 4. However, in recent years research has focused on the role of physical activity as a potential component in the prevention and/or management of depressive symptoms or clinical depression 1.

Several reviews examining observational and intervention studies have assessed the relationship between physical activity and depression/depressive symptoms 567. These reviews generally draw a similar conclusion: that physical activity is positively associated with reduced likelihood of depression or depressive symptoms. However, little is known about the specific components of physical activity that are important, particularly the dose, domain or social context of physical activity that might confer mental health benefits.

Whilst numerous studies have explored the associations between the dose – particularly the frequency and/or duration of physical activity – and likelihood of depression, findings from these studies were varied. A number of observational 8910 and intervention studies 1112 concluded that higher levels of physical activity (eg. greater than 60 minutes of physical activity per week) were associated with lower odds of depression. In contrast, other studies 13141516 have found that even low levels of physical activity (eg. exercising as little as 30 minutes per week) was related to improved mental health. Another study 11 has concluded specifically that the commonly recommended 'public health dose' of physical activity (at least 30 minutes of moderate intensity physical activity on most, if not all days of the week) 1718 was an effective treatment for reducing depressive symptoms.

Conflicting results were also found in studies investigating the association between physical activity intensity and depression. Numerous longitudinal 9, cross-sectional 19 and intervention studies 2021 have found that participating in physical activity at a moderate-intensity was significantly associated with lower odds of depression whilst few studies have examined the relative strengths of associations between physical activity of moderate or high intensity and risk of depression 2223.

Most studies supporting an inverse relationship between physical activity and risk of depression have focussed on leisure-time physical activity. Very little research has investigated the association of depressive symptoms with other domains of activity, such as work-related, transport-related or domestic physical activity. Only three studies have been identified which specifically compared leisure-time physical activity with physical activity in other domains 242526, with all three suggesting leisure-time physical activity was more strongly associated with lower odds of depression than physical activity undertaken in any other domain (domestic, transport-related, work). To our knowledge, no previous research has compared leisure-time, domestic, transport-related and work-related physical activity with odds of depressive symptoms in the same study.

It has been hypothesized that the associations between physical activity and depressive symptoms may be due to the social support and interaction that comes with physical activity undertaken in a group or social setting 27. Little research, however, has examined this hypothesis. Only one intervention study specifically compared physical activity undertaken alone with that undertaken with others 22. It found that physical activity had a moderate effect on alleviating depressive symptoms regardless of the social context in which it was undertaken.

The aim of this study was to examine the associations between various components of physical activity (dose, domain and social context) and odds of depressive symptoms in women using data from a large population-based cohort of women. Due to the cross-sectional nature of this study, only participants' current levels of physical activity and depressive symptoms were examined. Since a dose-response relationship has been previously reported 11, it was hypothesized that higher doses of physical activity (similar to that of the public health recommendations) would be more strongly inversely associated with odds of depressive symptoms than would lower doses. It was also hypothesized that leisure-time physical activity would be more strongly inversely associated with odds of depressive symptoms than other domains of physical activity (eg. work-related, transport-related and domestic), since activities performed during leisure-time are generally more enjoyable and undertaken by personal choice. A third hypothesis was that social physical activities and social support for physical activity would be more strongly associated with lower odds of depressive symptoms than physical activity performed alone or without social support.

The analyses in this study were based on cross-sectional survey data collected formerly from 1,554 women aged between 18 and 65. Full details of the methods are described elsewhere 2829.

Table 1 presents the socio-demographic characteristics and presence of depressive symptoms among participants. The mean age of participants was 42 years. The majority of the women were born in Australia (75%) and were married or in a defacto relationship (65%). A total of 615 (41%) participants reported their highest qualification as completing year 12 (or equivalent) or an apprenticeship or certificate/diploma, with 566 (37%) having completed a university or higher degree. Over one-quarter of the women were employed in a professional occupation (27%) and the majority of women surveyed did not have children living at home (60%). These demographic characteristics are comparable to those reported in other population-based studies conducted in Melbourne and nationally 45464748. A total of 421 (30%) of participants were classified as currently experiencing depressive symptoms (according to the GHQ-30). This is not dissimilar to previous findings which reported that almost 24% of Australian women had been diagnosed with depression 49.

Table 2 shows the proportion of women experiencing depressive symptoms according to sociodemographic and physical activity variables from chi-square analyses. The only domain of physical activity to be inversely associated with the presence of depressive symptoms was leisure-time (p = 0.009). Women who reported walking or moderate-intensity physical activity in leisure time were less likely to be experiencing depressive symptoms (p = 0.027 and p = 0.048 respectively). For example, 22% of those who reported more than 1.5 hours of moderate-intensity leisure-time physical activity (i.e. the highest tertile) were currently experiencing depressive symptoms, compared with 31% of those reporting no moderate-intensity leisure-time physical activity.

Being discouraged from physical activity and having no one to exercise with were associated with an increased presence of depressive symptoms (p = 0.002 and p = 0.037 respectively), whilst being a member of a sporting/recreational club, having someone to walk with in the neighbourhood and being active with a family member were associated with a lower presence of depressive symptoms (p = 0.044, p = 0.037, and p = 0.011 respectively).

Table 3 shows the crude and adjusted OR and 95% CI from logistic regression models predicting odds of depressive symptoms according to physical activity variables. Results from the unadjusted logistic regression models predicting odds of depressive symptoms according to duration of physical activity (summed across intensities) within each domain, showed that compared to those in the lowest tertile of leisure-time physical activity (i.e. <0.67 hours/week leisure-time physical activity), those women in the highest tertile (>3.5 hours/week) had lower odds of depressive symptoms (p = 0.002). Adjusting for covariates had little effect on this association. No other physical activity domain was associated with odds of depressive symptoms in logistic regression models.

When examined according to specific physical activity intensities, the unadjusted results showed that compared to those who reported no walking in leisure time, those in the highest walking tertile (>1.5 hours/week) had lower odds of depressive symptoms (p = 0.007). However, this association was only marginally significant when adjusting for confounders. Compared to those who reported no moderate leisure-time physical activity, adjusted results showed that those in the highest tertile (>1.5 hours/week) had lower odds of depressive symptoms (p = 0.039). Similarly, compared with those who reported no vigorous leisure-time physical activity, those in the highest tertile (more than 1.75 hours/week) had lower odds of depressive symptoms (p = 0.003).

There were no statistically significant associations between work-related physical activity of any intensity, and odds of depressive symptoms, in either unadjusted or adjusted logistic regression models. There were also no associations between total walking, moderate, vigorous or overall physical activity (summed across all domains) and odds of depressive symptoms.

Table 4 shows the crude and adjusted OR and 95% CI from logistic regression models predicting odds of depressive symptoms according to the social context of physical activity. The unadjusted results indicate that compared to those who reported never/rarely being discouraged from physical activity, those who reported sometimes being discouraged had higher odds of depressive symptoms (OR = 1.6; p = 0.005), while those who reported often being discouraged had even higher odds of depressive symptoms (OR = 2.4; p = 0.025).

These associations were only slightly altered (to OR = 1.7 and 2.3 respectively) in the adjusted analyses. Similarly in the unadjusted models, compared with those who reported being rarely/never active with a family member, those who reported being active a few times per year with a family member had lower odds of depressive symptoms (OR = 0.73; p = 0.022), while those who reported often being active with a family member had even lower odds of depressive symptoms (OR = 0.64; p = 0.009). These associations also changed only marginally (to OR = 0.77 and 0.61 respectively) after adjusting for covariates.

Compared to those who reported never/rarely having no-one to exercise with, the unadjusted results showed that those who often have no-one to exercise with had higher odds of depressive symptoms (p = 0.014). Although the adjusted results showed a similar trend, this was no longer statistically significant after covariates were included in the model. Being a member of a sporting club was significantly associated with lower odds of depressive symptoms before adjusting for confounders; however, this association only approached significance in the adjusted model (p = 0.061).

No association was evident between the frequency of having a friend/colleague to exercise with and odds of depressive symptoms, or having a dog to walk regularly and odds of depressive symptoms in either unadjusted or adjusted logistic regression analyses. Similarly, the adjusted results indicate that those who agreed that they have someone to walk with in the neighbourhood had lower odds of depressive symptoms, but this was not statistically significant (p = 0.083).

The hypothesized associations between overall physical activity dose and odds of depressive symptoms were not found in this study, as there were no significant associations between total walking, moderate, vigorous or overall physical activity (i.e. across all domains) and odds of depressive symptoms. These findings imply that it may not be the physiological effects of physical activity that are important in reducing the odds of depressive symptoms, since, if this were the case, a stronger association of odds of depressive symptoms with total dose (across all domains) of physical activity would have been expected.

The only domain of physical activity found to be inversely associated with odds of depressive symptoms was leisure-time activity. When controlling for age, marital status and physical health, those who reported more than 31/2 hours of total leisure time physical activity per week (summed across intensities) had lower odds of depressive symptoms when compared to those who undertook less than this. In fact, no significant associations were found between undertaking less than 3 1/2 hours of total weekly leisure-time physical activity and odds of depressive symptoms, suggesting that a high duration of leisure-time physical activity may confer greater mental health benefits. This finding supports the dose of physical activity recommended in the US, UK and Australian national physical activity guidelines which suggest a minimum duration of physical activity equivalent to 2 1/2 to 3 1/2 hours per week (at least 30 minutes on most, if not all days of the week 171850. This is consistent with findings from several observational 851 and intervention studies 5253 which found the public health dose of exercise to be strongly associated with a lower risk of depression. The Australian national physical activity guidelines are based largely on evidence concerning benefits for physical health. These findings suggest they could be extended to mental health. Although other studies have also found a shorter duration of physical activity to be associated with lower risk of depression 141954, these studies used measures of physical activity and risk of depression that differed from the current study. For example, unlike the current study, the method used by Hassmen et al 19 was unable to determine different intensities or domains of physical activity.

When examining different intensities of leisure-time physical activity, the results showed reduced odds of depressive symptoms amongst women in the highest tertiles of moderate- and vigorous-intensity leisure-time physical activity, but those performing vigorous-intensity physical activity showed an even greater reduction in odds. These findings are supported by previous studies that found a strong association between undertaking higher durations of moderate-intensity 9 or vigorous-intensity 51 physical activities and odds of depression. Findings from the current study suggest that the greater the dose (intensity and duration) of leisure-time physical activity, the lower the odds of depressive symptoms. However, there was also an inverse trend approaching significance between depressive symptoms and walking, suggesting that vigorous-intensity physical activity may not be necessary to achieve the reduced odds of depressive symptoms associated with physical activity. Targeting physical activity of specific intensities perhaps guided by personal preference and enjoyment may be a point of consideration when prescribing or recommending a dose of physical activity to confer mental health benefits, as this may result in an increased likelihood of maintenance of the training program and consequently maintenance of reduced depressive symptoms 55. However, given there were strong inverse associations of depressive symptoms with vigorous activity, women should be encouraged to include at least some activity of this intensity in their weekly physical activity routine if possible.

Several researchers have hypothesised that physical activity reduces risk of depression through physiological pathways – for instance, exercise may activate endorphin secretion, which reduces pain and produces a euphoric sensation 1. However, the finding that leisure-time physical activity was the only domain associated with odds of depressive symptoms suggests some alternative possible explanations, such as a sense of enjoyment or a perceived control or choice when undertaking activities in leisure-time that may be lacking when physical activity is undertaken in domains other than leisure-time (i.e. work-related, domestic, and transport-related). Alternatively, the finding that different domains of physical activity were differentially associated with odds of depression could be attributed to recall difficulties, as leisure-time physical activity may be more accurately recalled than physical activity undertaken in other domains 56. This finding is in contrast to previous observational studies 2425 that did find weak correlations between physical activity undertaken in other domains (domestic and transport) and likelihood of depression. However, it was difficult to interpret the results from those studies in terms of optimal domain of physical activity for reducing likelihood of depression, since in these studies, domain was simply equated with intensity (eg. domestic physical activity = low-intensity physical activity), yet physical activities undertaken from the domestic domain are not always of a low intensity, and not all low-intensity physical activity is necessarily domestic in nature 57. Few studies have examined the association between multiple domains of physical activity and likelihood of depression. However, concurrent with findings in this study, one previous observational study that compared leisure-time physical activity with domestic and work-related physical activity, found only leisure-time physical activity to be inversely associated with depression 26.

The findings of the current study are generally consistent with the social interaction hypothesis, which posits that the improvements in mental health following exercise are at least partly related to the mutual support and social relationships that are provided when participating in physical activity 27. In the current study, lower social support for physical activity (eg. being discouraged by others) was associated with higher odds of depressive symptoms, whilst greater social support (eg. being often physically active with a family member) was associated with lower odds of depressive symptoms. These findings, however, are not consistent with one previous study of men and women that suggested that physical activity was associated with lower odds of depressive symptoms regardless of the social format 22. The sample in that study differed from the current study (i.e. included both men and women), which may explain the contrasting results; women may value and benefit more from the social aspects of physical activity more than men 58. In the present study, no association was seen between being physically active with a friend/colleague, having no-one to exercise with or having a dog to walk regularly and odds of depressive symptoms. It may be that the support for being physically active from family is more important for mental health among women than support from other sources.

Several limitations of this study should be acknowledged. Firstly, the cross-sectional nature of the study did not allow us to determine the causality or direction of relationships. For example, those without depressive symptoms may be more capable and motivated to participate in physical activity. We were also unable to determine whether the association between support for physical activity and depressive symptoms is due to the social aspect of the physical activity, or due to having a generally more supportive social network. Further longitudinal and intervention studies are needed to confirm the associations observed here, and also to investigate the mechanisms by which physical activity might protect against risk of depression. Such hypothesized mechanisms include 'distraction', whereby improvements in mental well-being following exercise are due to diverting negative thoughts and unpleasant stimuli during the activity 1; or 'mastery', whereby improvements following physical activity are due to achieving goals and a feeling of success 1. This study relied on self-report measures which possess the potential for error in judgement, recall difficulties and the possibility of socially desirable responses. Nevertheless, measures of depressive symptoms, physical activity and social influences were valid and widely accepted 353738394042. Finally, the response rate for the physical activity survey was only 43.5%, whilst the response rate from those who received the diet questionnaire first was 21.2%. This is a potential limitation of our study as participants who responded may be more interested in physical activity and possibly be more active than non-respondents. However, comparison of physical activity levels of the women in this study with those of Australian women reported elsewhere 46 showed similar levels of activity across the two samples.

Strengths of this study include the large, population-based sample of participants who were generally representative of Australian women in terms of both their physical activity levels and depressive symptoms 4649. Further, to our knowledge, no previous research has compared leisure-time, domestic, transport-related and work-related physical activity with odds of depressive symptoms in the same study, and few have examined the social context of physical activity and its association with odds of depressive symptoms. Therefore, this study provides a novel perspective on the optimal dose, domain and social context of physical activity that may confer mental health benefits in women.

In summary, acknowledging the need for confirmation of causal effects, the results from this study suggest that promoting a relatively high duration of leisure-time physical activity could be an important strategy in the prevention of depressive symptoms in women, with additional benefits potentially resulting at higher intensities. Furthermore, programs aimed at reducing the odds of depressive symptoms through the promotion of physical activity could consider the potential for additional benefits associated with physical activity undertaken in a social context by developing strategies emphasizing social aspects of leisure-time physical activity, such as walking groups or exercise classes.

The authors declare that they have no competing interests.

MT performed the analyses and led the writing of the manuscript. KB conceived of the study, participated in the design, the survey development and helped to draft the manuscript. JS contributed to the survey questionnaire and helped to draft the manuscript. All authors read and approved the final manuscript.