Exercise and depression
Mental illness is a significant impediment to health and well-being (Department of Health 1999). Reducing mental illness and improving mental health services is a milestone of the UK Government's health service reforms reflected in the National Service Frameworks for Mental Health issued in 1999. In 1987 the US National Institute Mental Health (NIMH) assembled a panel of experts to produce a consensus statement on the mental health effects of exercise and thus reconcile research and clinical practice. The resultant consensus statement concluded that exercise is (i) positively linked with mental health and well-being, (ii) reduces stress and state anxiety, and (iii) has emotional benefits for all ages and in both genders (Morgan & O’Connor 1988). A recent World Health Report (WHO 2000) shows that depression is a significant mental health problem afflicting people living in all member states. Among males it is ranked eighth accounting for 2.8% of Disability Adjusted Life Years (DALY’s). Among females it is ranked third accounting for 5.8% of DALY’s. Running is advocated as a more effective antidote to depression than psychoanalysis (Greist et al. 1979). Empirical and anecdotal evidence show that exercise may have an antidepressant effect in healthy individuals (North et al. 1990, DiLorenzo et al. 1999), and among those with profound multiple disabilities (Green & Reid 1999). Schizophrenia is one of the most disabling of all mental illnesses. Exercise is shown to reduce auditory hallucinations, raise self-esteem, and improve sleep patterns and general behaviour in people living with schizophrenia (Faulkner & Sparkes 1999).
North et al. (1990) reviewed the results of narrative and meta-analytic reviews investigating the effect of exercise on depression. In this review exercise is suggested to improve depression by changing people's daily routine, increasing their interactions with others, helping them lose weight, participate in outdoor recreation and master difficult physical and psychological challenges. Evidence that biological factors may explain the beneficial effects of exercise on depression derives from research showing that exercise promotes the secretion of neurotransmitters like serotonin (Ransford 1982, Morgan 1985). Also, evidence from animal studies suggests exercise stimulates the secretion of endogenous morphines (‘Endorphins’) and produces a state of euphoria (Pert & Bowie 1979). The narrative reviews reviewed by North et al. (1990) provide evidence for the benefits of exercise on depression but these derived mainly from anecdotal observations. The meta-analytic review, however, supported the anecdotal observations. The effect of acute exercise (single exercise session) was different than for an exercise programme, but both were effective antidepressants. Exercise had a better effect on outcomes for respondents who were most physically and psychologically unhealthy at the outset of the studies. A particular strength of this meta-analysis is that the authors analysed the results of published and unpublished studies. This was important because respondents in published studies had lower levels of depression when compared with respondents in unpublished studies. Another strength of this meta-analysis is that the authors’ accounted for factors that may have produced the positive effects on depression attributed to exercise, e.g. source of participants, purpose of exercise, exercise location, group assignment of participants, age, gender and mode and duration of exercise. Effect sizes (the size of the difference in outcomes between two interventions) for exercise were higher when respondents performed exercise for medical rehabilitation, at home, when they were randomly assigned to exercise conditions and when the exercise was of medium intensity. Middle-aged males undergoing haemodialysis benefited most from exercise. Weight training of between 21 and 24 weeks produced lower depression levels.
On the basis of a systematic review of the effect of exercise on depression as measured by Beck Depression Inventory (BDI) scores, Lawlor & Hopker (2001) reported that exercise produced a large decrease in depression symptoms (effect size = 1.1) when compared with no treatment. However, most of the studies in this review were of poor quality (inadequate randomization and lack of blinding in outcome assessment), had brief follow-up and sampled non-clinical volunteers. Schulz et al. (1995) have shown that such methodological weaknesses exaggerate results in favour of the intervention by 20–40%. Lawlor & Hopker (2001) thus conclude that the effectiveness of exercise on depression is undetermined, a conclusion challenged by others. Mutrie (2002) argued that the large effect size reported by Lawlor and Hopker offers compelling evidence for the benefit of exercise and points out that unlike drug therapy, exercise has few negative side-effects. In a letter to the British Medical Journal (BMJ, 2nd April 2001) commenting upon the Lawlor and Hopker review, Biddle (2001) suggested that the exclusion of other types of studies in the original review, such as large population surveys led the authors to underestimate the benefits of exercise. It is hard to disagree with Biddle especially in light of arguments Lawlor raises in a reply to Biddle's letter (BMJ, 10th April 2001). Lawlor argues that the statistics reported by Schulz et al. (1995) showing methodological weaknesses in trials exaggerate the effects of interventions by between 20 and 40% (see above) and thus invalidate the evidence for exercise on depression. On closer scrutiny it would be more accurate to conclude, in my view, that the evidence is weakened but not invalidated. In their review Lawlor and Hopker report an average effect size for exercise on BDI scores of 1.1. Taking the upper end of Schulz et al.'s figure, we can reduce this effect size by 40% to account for methodological weaknesses. This leaves an effect size of 0.66, medium by widely cited estimates (Cohen 1992). Statistically, those who exercised were 0.66 standard deviations less depressed and scored 7.3 points lower on the BDI than those who did not exercise. It is difficult to interpret what this means exactly for clinicians treating depression, or people experiencing depression. But if we consider how the BDI is used in clinical practice as an indicator of the experience of depression the statistics may have significant clinical relevance.
For example, Mary Blue (pseudonym) is enrolled on an exercise on prescription programme to treat her depression. Prior to starting the exercise regime, Mary has a BDI score of 24 indicating moderate to severe depression (Stinton & Devilly 2002). Following the exercise programme, Mary's BDI score drops to 16.7 indicating mild to moderate depression. Mary is significantly less depressed and more likely to re-engage with her activities of living, a good result by anyone's reckoning. Psychotropic drugs with harmful side-effects have been prescribed liberally with smaller effect sizes than exercise. Given the difficulty in translating outcomes from studies using continuous variables such as BDI scores in clinical practice, Lawlor & Hopker (2001) exhort researchers to use more dichotomous (i.e. depressed/not depressed) outcome measures. This is a clinically naïve position; depression is a question of degree and is seldom, if ever, dichotomized so easily. The BDI is after all a widely used measure of depressive features with consistent validity (Richter et al. 1998).
It is important to note that no evidence of effect as found by Lawlor and Hopker does not mean that no effective evidence exists. In the most recent issue of Clinical Evidence – the so-called international source of the best available evidence for effective health care –Geddes et al. (2003) state that exercise has limited effectiveness in the treatment of depression. This view conflicts with the National Quality Assurance Framework for Exercise (NQAFE) (Department of Health 2001) that states exercise has causal impact on mental health problems including depression. Also, the National Consensus Statements on exercise and mental health published by the Health Education Authority (Grant 2000) state that exercise is consistently associated with positive affect, mood and psychological well-being.
Exercise and anxiety
Petruzzello et al. (1991) conducted three meta-analyses to examine the effect of acute and chronic exercise on state (current) anxiety, trait (dispositional) anxiety and psychophysiological correlates of anxiety derived from published and unpublished studies (n = 104) reported between 1960 and 1989. The results for state anxiety showed that exercise produced a small effect on state anxiety (effect size = 0.24); the effect size was largest when the researchers used the Multiple Affect Adjective Checklist (MAACL) as the outcome measure of anxiety. Chronic exercise had a slightly better effect on anxiety than acute exercise. The effect of exercise was largest in pre–post test within-groups designs, aerobic exercise was better than anaerobic exercise, and high intensity exercise of 21–30 min duration had a better effect than low intensity exercise shorter than 20 min or longer than 30 min. The effect size was largest when anxiety was measured 20 min post-exercise. Effect sizes were largest in studies using matched controls and lowest in studies using random assignment. Effect sizes were largest in participants aged between 31 and 45.
When trait anxiety was the outcome measure the effect size for exercise was moderate (0.34) overall. High intensity aerobic exercise of more than 40 min duration, performed for more than 15 weeks produced the highest effect sizes. Effect sizes were largest in participants aged below 18. For trait anxiety, exercise had the largest effect in people with a psychiatric illness.
When psychophysiological correlates of anxiety were the outcome measures, exercise had a fairly large effect on anxiety (effect size = 0.56) overall. The effect size was largest when skin measures were used to measure anxiety. Acute exercise had a better effect on psychophysiological outcomes than chronic anxiety. Pre–post within-groups designs produced larger effect sizes than other designs. Lower intensity exercise, of up to 20 min duration, lasting 4–6 weeks produced the largest effect sizes. Effect sizes were largest in studies using matched controls, among 18 to 30-year-olds and those with a psychiatric illness.
The results from these detailed meta-analyses show that aerobic exercise is associated with reductions in anxiety although the effects were not uniform across the three outcome measures. Exercise fared similarly to other interventions (e.g. relaxation) on trait anxiety. Random assignment was needed for effects on trait anxiety as were programmes exceeding 10 weeks. The variable that was significant across all outcome measures was duration of exercise; exercise of more than 20 min duration appears necessary for reduction in anxiety levels, irrespective of how anxiety is measured. However, the effect of exercise on anxiety was more apparent in studies eschewing randomization and not controlling for the effect of other variables on anxiety. When the studies were better designed, the effect for exercise was smaller.
In a narrative review of the effect of exercise on psychological health, Weyerer & Kupfer (1994) examined data from observational studies as well as controlled trials. The authors concluded that exercise improved psychological health, in some cases better than counselling alone, even when controlling for unspecified sociodemographic and health-related confounding variables.
There are several views that seek to explain the beneficial effects of exercise on anxiety. One view suggests that exercise raises body temperature and reduces muscle tension similar to the effect of having a warm bath – the so-called thermogenic hypothesis (Raglin & Morgan 1985). Another view suggests that exercise stimulates activity in the sympathetic nervous system (SNS); adrenaline levels are increased and this has an arousing effect. When the SNS is activated, it provides a catalyst for parasympathetic nervous system (PNS) activity; acetylcholine is released and this has a calming effect. This is known as the Opponents Process Model (Solomon 1980). Exercise is also thought to distract people from stressful events thereby reducing the anxiety provoking impact of these events (Bakre & Morgan 1978). It is unclear from the studies reviewed by North et al. (1990) how long term are the effects of exercise on anxiety. A recently published study addressed this issue.
DiLorenzo et al. (1999) investigated the effects of exercise on self-reports of depression, anxiety and self-concept and aerobic fitness, heart rate and maximum oxygen uptake. Eighty-two participants aged between 18 and 39 were randomly allocated to a 12-week programme of bicycle ergometry or a control condition and followed up 12 months later. At the end of the programme and at follow-up participants allocated to the exercise programme had more positive changes in all outcomes than the control participants.
It is rare to find studies investigating the role of exercise on health that use qualitative research methods. One exception is the study reported by Faulkner & Sparkes (1999). Using an ethnographic design – participant observation and interviews with participants and their key-workers – Faulkner and Sparkes explored the therapeutic value of a 10-week exercise programme of twice weekly sessions, for people living with schizophrenia. The authors found that exercise reduced participants’ perception of auditory hallucinations, raised their self-esteem and improved their sleep patterns and ‘general behaviour’. Exercise provided distraction and social interaction and these accounted for the exercise benefits, the authors concluded. It may be that the three individuals studied in this investigation are unrepresentative of the experiences of people living with schizophrenia. Also, it is unclear from this study how exercise interacted with other interventions, like medication, that may have produced the positive outcomes. Nevertheless, the ethnographic design allowed the researchers to study intensively the participants and their key-workers’ experiences. The results of this study support the therapeutic value of exercise. The results therefore converge with the findings of many quantitative studies and this convergence lends further support to researchers’ claims that exercise improves mental health.
Exercise and cognitive functioning
Cognitive functioning is an important part of mental health and well-being and researchers have used cognitive functioning as an outcome measure in many studies testing the effects of exercise on mental health.
Van Sickle et al. (1996) conducted a meta-analysis of 18 previously published studies that examined the effects of exercise on cognitive functioning in elderly people. Exercise interventions used in the studies reviewed in this meta-analysis included bicycle ergometry, walking and callisthenics, flex and stretch exercises, aerobic workouts of varying levels of intensity and duration and non-aerobic stretch and coordination exercises. Outcome measures used in the studies in this meta-analysis included performance on memory tests, mathematical tests, IQ tests and perceptual organisation. The studies included in this meta-analysis included pre–post test within-group designs and pre–post test between-group designs using either two conditions, or three or more conditions. Many studies used multiple outcome measures. The researchers analysed the findings from the original studies in three ways: (1) a comparison of the number of statistical tests reporting significant findings with the total number of statistical tests used, (2) a summary of the findings for specific outcome measures, and (3) an examination of the effect sizes from each study. The results show 34% of the statistical tests used yielded significant findings; exercise promoted improved cognitive functioning. Of the 66% of studies yielding non-significant findings 30% had such small sample sizes, that there was a high risk of type 2 errors (power = 0.05–0.43). Only mathematical abilities were significant across studies. By Cohen's (1992) estimate, 11 effect sizes were small (0.01–0.18), 20 were small to medium (0.21–0.48), 9 were medium to large (0.50–0.77) and 6 were large (0.80–1.59). Exercise produced a moderate improvement in cognitive functioning. Yoga produced the largest effect sizes for most outcome measures. Many of the studies using aerobic exercise produced non-significant findings but the statistical power of these studies was very low. Thus, aerobic exercise may have had significant effects but many of the studies using this intervention had insufficient sample sizes to detect these effects. When the researchers excluded these studies from the meta-analysis, the percentage of studies showing significant effects rose from 34% to 47%. Elderly adults often show poorer cognitive function when compared with younger adults, especially when their performance is measured on IQ tests (Van Sickle et al. 1996). Therefore, perhaps the cognitive benefits attributed to exercise shown in this meta-analysis only emerge in people whose cognitive function is below optimal levels. The meta-analysis conducted by Etnier et al. (1997) addresses this issue as it included studies sampling participants of most age groups.
Etnier et al. meta-analysed the results of 134 studies that investigated the effects of exercise on cognitive functioning. Participants’ ages in this analysis ranged from 6 to 90. Exercise interventions used in the original studies included short and long-term applications of aerobic and anaerobic exercise, muscular resistance, callisthenics and step tests, of various intensity. The location of the interventions included home, laboratory, hospital, fitness centre and classroom. Outcome measures included line matching tests, motor skills test, verbal comprehension tests, the Stroop test, Ravens Progressive Matrices IQ test, the Weschler Memory Scale and the Sternberg number task. The results showed that the overall effect size for exercise on cognitive functioning was 0.25, small by Cohen's (1992) estimates. More detailed analyses revealed that a mixed (acute and chronic) exercise intervention produced the largest effect size (0.54). When the researchers analysed the effect of the acute exercise interventions separately studies using random sampling produced the largest effect size (0.65), but so did studies that had more threats to internal validity (1.76). Exercise had the largest effect on motor-skills performance (1.47), in studies using samples of both genders (0.70) and studies using exercise groups of 20 or more people (0.61). When the researchers analysed the effect of the chronic exercise interventions separately, studies using a single group design had the largest effect size (0.88), as did studies that had more threats to internal validity (0.57), those using older adults (45–60 years, effect size = 1.02), in studies where exercise was conducted in a classroom (effect size = 0.67), as did studies using exercise groups with 10 people or fewer (1.22). When the researchers analysed the data from cross-sectional/correlation studies separately low intensity exercise had the largest effect size (1.12), as did studies using exercise groups of fewer than 10 people (0.85), and when the researchers did not report the exact outcome measure (1.69). Overall, chronic exercise produced a larger effect size (0.33) than acute exercise (0.16).
In this meta-analysis as effect size increased threats to internal validity increased; thus, the poorer the study, the larger the effect size. This suggests that the link between exercise and cognitive functioning may be more correlation and less causation. However, a closer examination of the effect sizes when well-designed studies with few threats to internal validity are considered separately shows a small (0.18), but significant effect on cognitive functioning. The results of this meta-analysis also suggest that the benefits of exercise on cognitive functioning are more marked in older adults. This meta-analysis, like all other studies and meta-analyses, show the effect sizes for exercise on various outcome measures. However, none of the studies or meta-analyses explain what, exactly, the effects sizes mean in terms of longevity and general quality of life. Curfman (1993) estimates that regular intensive exercise may add up to 2 years on life expectancy.
In summary, there is consistent evidence from many studies using various research methods demonstrating that aerobic exercise of 20–30 min duration performed between three and five times per week is beneficial to mental health and well-being. There are, however, consistent methodological limitations in many of the exercise studies. These limitations weaken the case for exercise, but they do not invalidate it. The final part of this paper examines whether mental health practitioners use exercise as a therapeutic intervention and provides guidance on how exercise can be incorporated into the routine care of people with mental health problems.