Article Text
PDF
Abstract
Aim Evidence for a link between physical activity and mental health in young people is hampered by methodological shortcomings. Using repeat assessments of device-measured physical activity, we examined the association of within-individual variation in free-living activity over 7 years with depressive symptoms.
Methods This was a prospective cohort study of a nationally representative sample of children born in the UK (n=4898). Physical activity was quantified using accelerometry at ages 7 and 14. The main outcome was depressive symptoms, based on the Short Mood and Feelings Questionnaire, assessed at age 14.
Results After adjustment for socioeconomic status, body mass index and psychological problems at baseline, a higher level of light-intensity activity at age 7 in girls was associated with a lower likelihood of having depressive symptoms at follow-up (OR, 0.79; 95% CI 0.61 to 1.00), although no associations were observed for moderate to vigorous activity or sedentary behaviour. Girls who transitioned from low baseline activity to higher levels at follow-up experienced a lower risk of depressive symptoms (OR, 0.60; 95% CI 0.39 to 0.92) compared with the inactive reference category. Null associations were observed in boys. Participants who consistently met the current recommendation of 60 min/day of moderate to vigorous activity both at 7 and 14 years of age experienced the lowest risk of depressive symptoms (OR, 0.55; 95% CI 0.34 to 0.88).
Conclusion To prevent depressive symptoms in adolescence, policies to increase physical activity from mid-childhood may have utility.
- child health
- cohort studies
- depression
- physical activity
Statistics from Altmetric.com
Introduction
Mental disorders, including depression, account for a large proportion of the burden of disease in young people worldwide.1 2 As well as being an important public health concern in their own right, mental illness in young people impacts on a range of later outcomes, including educational achievements, illicit drug use, self-harm, sexual health, life expectancy and chronic somatic disease.1 3–5 In many individuals, the first onset of depression occurs in adolescence,6 and even subclinical symptoms during this period are associated with increased risk of adulthood disorder,7 magnifying the need to understand and prevent experience of moderate levels of distress. While evidence-based treatments exist,2 these are not universally effective, raising the need to identify modifiable, policy-relevant risk factors. Behavioural strategies such as physical activity (PA) may therefore be a particularly important approach in the prevention of depression.
A series of observational studies suggest that higher levels of PA are associated with a lower risk of depressive symptoms in younger people.8 9 While informative, these studies are subject to several methodological constraints that hamper data interpretation. A major concern is the preponderance of cross-sectional studies,8 a methodological design that is subject to reverse causation, whereby physical inactivity may be linked to poor psychological health but the converse may also be the case. A second issue concerns the time-varying nature of PA, particularly during the transition from childhood to adolescence,10 that a single assessment of PA will fail to capture. Third, the near-exclusive use of self-reported PA measures in these studies raises well-documented concerns regarding response bias11 and hampers the ability to explore PA across the full behaviour spectrum, including sedentary time.
Some of these issues can be overcome by the use of randomised controlled trials12 that employ non-exercise control groups. However, that PA may be beneficial for many dimensions of health means that the long-term encouragement of sedentary behaviour is ethically unacceptable. An alternative approach is to use an observational study setting13 with repeat assessments of PA, whereby changes in activity are tracked and related to the occurrence of psychological problems. If data on change in PA were also objectively gathered on repeat occasions, as opposed to being ascertained via self-report,14 then in addition to increasing confidence in the results the full range of free-living activity would also be captured, including sedentary behaviour, which may exert its own influence on health. Repeat assessment of device-measured PA in sufficiently large cohorts is, however, rare. For the first time, to our knowledge, we used data from a large-scale, nationally representative prospective cohort study with repeat assessments of device-measured PA to examine the association of changes in free-living PA over a 7-year period with depressive symptoms.
Methods
Participants
A nationally representative sample of children born in the UK were recruited into the Millennium Cohort Study between September 2000 and January 2002. The study has been described in full elsewhere.15 In brief, eligible children were identified from child welfare benefit records, a scheme covering nearly all families in the UK. To date, there have been six waves of data collection (at age 9 months and at ages 3, 5, 7, 11 and 14 years). We used the fourth wave, when participants were aged 7, as the baseline for the present study as this was the first occasion that device-measured PA data were gathered. Interviewers visited the cohort members’ homes and conducted face-to-face interviews with both parents.
PA assessment
During the fourth wave of data collection, when participants were aged 7 (2008/2009), PA and sedentary time were measured over a 7-day period using the waist-worn uniaxial ActiGraph GT1M accelerometer (ActiGraph, Pensacola, Florida, USA). Sedentary time was defined as <100 counts per minute, light PA (LPA) as 100–2241 and moderate to vigorous (MVPA) as >2241.16 PA assessments were repeated at age 14 using a triaxial accelerometer (GENEActiv, Activinsights, Cambs, UK) placed on the non-dominant wrist for two full 24-hour period on a randomly selected week and weekend day.17 Data were downloaded using GENEActiv software and raw data processed using the GGIR package in R (https://cran.r-project.org/web/packages/GGIR/GGIR.pdf),18 which includes autocalibration and non-wear detection functions. The principal output was the Euclidean norm minus one (ENMO), which is a measure of mean acceleration over a 24-hour period from three axes (sagittal, frontal and vertical) relative to the horizontal plane.
Depressive symptoms
At baseline, when study members were 7 years of age, psychological distress was assessed using parental reports from the Strengths and Difficulties Questionnaire (SDQ), a 25-item tool using 3-point Likert scale which has demonstrated good reliability and validity for use in population-based surveys.19 The SDQ was originally validated on a nationwide sample of 10 438 British children aged 5–15 years old, demonstrating internal consistency (mean Cronbach’s α=0.73), cross-informant correlation (mean: 0.34) and retest stability after 4–6 months (mean: 0.62).19 A total difficulties score was derived by summing the subscales of hyperactivity, emotional symptoms, conduct problems and peer problems, with a higher score denoting greater distress. At follow-up, when cohort members were aged 14, the Short Mood and Feelings Questionnaire (SMFQ) was used.20 Consisting of 13 items, a score of 12 points or higher indicates greater depressive symptoms. In the present cohort, the SMFQ demonstrated strong internal reliability (Cronbach’s α=0.93). Several studies have reported moderate to high criterion validity for discriminating children aged 7–17 years old with and without major depressive episodes using the Diagnostic Interview Schedule for Children as the criterion standard.21 22
Covariates
Covariates for the present analyses were selected a priori based on existing evidence,8 9 and included parental social occupational class (using the Registrar General’s classification: professional and managerial occupations; skilled, non-manual occupations; skilled manual occupations; and routine and manual occupations) and body mass index (BMI: weight in kg/height in m2) as calculated from baseline height and weight. Height was taken using a Leicester height measure stadiometer with a Frankfurt plane card. Weight was measured using Tanita scales (BF-522W) to the nearest 0.1 kg. Although BMI may be viewed as being on the intermediate pathway between PA and depression, we made an a priori decision to model BMI as a confounder based on previous studies.8
Statistical analysis
We carried out two sets of analyses. In the first, we related baseline assessment of PA to depressive symptoms 7 years later, and in the second we explore the association of change in PA between 7 and 14 years of age with subsequent depressive symptoms. In analyses where we categorised depressive symptoms at follow-up, we used logistic regression to compute OR with accompanying 95% CI to summarise the relation with sedentary behaviour, LPA and MVPA at baseline; where a continuous score was used for depressive symptoms, linear regression produced beta coefficients. In these analyses we adjusted for parental socioeconomic category based on occupation, BMI (continuous variable), device wear time and total SDQ score at baseline. The relation of MVPA and LPA with depressive symptoms was mutually adjusted, while sedentary behaviour was adjusted for MVPA only.23 Second, to compute PA change between ages 7 and 14, we used continuous counts per minute and ENMO mean acceleration data at baseline and follow-up, respectively, to categorise participants into ‘high’ or ‘low’ activity based on the sex-specific median split. We then created four activity groups based on PA measurement at both time points (low–low; low–high; high–low; high–high). We have previously found evidence of selection bias in those participants providing valid accelerometry data24; thus, we used weighted analyses based on the accelerometry subsample. Given the marked gender differences in depressive symptoms and PA during the adolescent developmental period, all analyses were stratified by sex. We also explored the impact of reverse causality by removing from the analyses study members regarded as a ‘case’ at baseline (SDQ score >16). All analyses were conducted using SPSS V.22, with statistical significance set at p<0.05.
Results
Based on the inclusion criteria (at least 2 days with ≥10 hours of wear time),16 6675 study members (3176 boys) provided viable accelerometry data at baseline (online supplementary table S1). After exclusion of those with missing covariates and follow-up data, the analytical sample comprised 4763 participants. Excluded participants were more likely to come from more socioeconomically deprived families (semiroutine/routine occupation: 33.1% vs 23.6%, p=0.001) and had higher mental health problems (SDQ score at baseline: 7.5±5.3 vs 6.2±4.6, p=0.001) than those included, although there were no differences in PA (605±161 cpm vs 601±148 counts per minute (cpm), p=0.24) or BMI (16.4±2.3 kg/m2 vs 16.4±2.0 kg/m2, p=0.59).
Supplemental material
PA at 7 years of age and depression at 14 years
In table 1 we show the characteristics of study members according to baseline PA. Participants in the highest PA tertile at baseline were much more likely to be boys, but socioeconomic status and BMI were similar across the PA groups.
Baseline characteristics of study members according to baseline physical activity (n=4763)
At follow-up, 16.4% of the sample were regarded as a case based on their depressive symptoms. Baseline LPA was inversely associated with later depressive symptoms in girls but not in boys. No associations were observed for MVPA or sedentary behaviour (table 2). With there being a moderate correlation between LPA and sedentary time (r=0.35), we reran the analyses making mutual adjustment. The association between LPA and depressive symptoms was largely unchanged (highest tertile of LPA, OR=0.80; 95% CI 0.61 to 1.05), although associations with sedentary and depressive symptoms were entirely attenuated after adjustment for LPA (highest tertile of sedentary, OR=1.00; 95% CI 0.75 to 1.34). To explore reverse causality, we repeated the analyses after removing 6.7% of the sample displaying psychological distress (SDQ scores >16) at baseline. The pattern of results was essentially the same (online supplementary table S2). For example, in girls there were lower odds of elevated depressive symptoms in the highest tertile of LPA (OR=0.78; 95% CI 0.60 to 1.00), and null associations remained for sedentary and MVPA.
Prospective association between physical activity at baseline and depressive symptoms at age 14 stratified by sex
Change in PA and depressive symptoms
Change in PA was calculated for a sample of 2416 participants who had available device-measured PA data from ages 7 and 14. The four activity groups were generally evenly distributed, and participant characteristics are shown in online supplementary table S3. A total of 1063 (44%) children in this sample changed PA category between baseline and follow-up. There was evidence that, relative to those children who remained inactive over the observation period, participants who transitioned from low baseline activity to higher activity at follow-up had lower odds of depressive symptoms, particularly in girls (OR=0.60; 95% CI 0.39 to 0.92) (table 3). Participants who were less active at follow-up had the same risk of having depressive symptoms as children who were inactive throughout. The results were largely replicated when we categorised participants based on meeting the current PA guidelines (60 min/day MVPA) using thresholds to define MVPA (>2241 cpm on ActiGraph16; >100 mg on GENEActiv25). Participants who met the PA guidelines both at baseline and follow-up displayed the lowest odds of depressive symptoms (OR=0.55; 95% CI 0.34 to 0.88), as did participants who met the guidelines at follow-up but not at baseline (figure 1).
Association between change in activity from ages 7 to 14 and depressive symptoms at age 14 stratified by sex
Association between change in activity from ages 7 to 14 and depressive symptoms at age 14 (n=2416). aContinuous counts per minute and mean acceleration data at baseline and follow-up, respectively, were used to categorise participants into ‘high’ or ‘low’ activity at ages 7 and 14 based on the sex-specific median split. bA threshold of 60 min/day of MVPA was used to categorise participants meeting the PA guidelines: ‘yes’ or‘no’. Effect estimates are adjusted for sex, total SDQ score at baseline, parental occupational social class, body mass index at baseline and PA at baseline (cpm). CPM, counts per minute; MVPA, moderate to vigorous physical activity; PA, physical activity; ref, reference; SDQ, Strengths and Difficulties Questionnaire; SMFQ, Short Mood and Feelings Questionnaire.
We repeated the analyses using the continuous PA data. The results showed that higher PA (per SD mean acceleration) by age 14 was inversely associated with depressive symptoms (OR=0.53; 0.33–0.86) after adjustment for PA at baseline.
In sensitivity analyses the associations of PA change and depressive symptoms were not changed when we removed participants displaying psychological distress (SDQ scores >16) at baseline (online supplementary table S4).
Discussion
Our results show an association of baseline LPA and increases in PA between childhood and adolescence with lower risk of depressive symptoms. Associations were observed in girls but not boys. A key feature of the present study was the repeat assessments of device-measured PA, which allowed us to examine the effects of within-individual free-living PA variation over 7 years on depressive symptoms. Indeed, associations between MVPA and depressive symptoms only emerged when we used data at ages 7 and 14, highlighting the importance of collecting repeat data. In randomised controlled trials of paediatric populations, it would be unethical to assign participants to a control group instructed to maintain very low levels of PA for a prolonged period of time. Indeed, the use of appropriate control groups in this field has been a source of debate and possible bias.12 26 Further, free-living activity measures are ecologically more informative than highly structured, often supervised training regimens typically found in trials, which often yield high dropout, raising concerns regarding long-term sustainability.
Study limitations
We cannot discount the possibility of residual confounding due to unmeasured or superficially assessed covariates. For example, physical illness may have driven both inactivity and depressive symptoms, although participants of the present sample were largely healthy. Also, while our outcome was depressive symptoms rather than diagnosis, the SMFQ has demonstrated acceptable criterion validity.21 22 Parental rather than self-report of psychological distress in childhood is a potential source of bias. As anticipated,24 participants included in our analytical sample were more socioeconomically advantaged than those who dropped out. Weightings were used to partly overcome selection biases, although exclusion of more deprived cohort members with greater mental health issues may have partially diluted our findings. While device-measured PA overcomes some of the biases associated with self-reported PA,11 this approach only provides data for a small time window in participants’ lives, and this may not be a true reflection of their habitual behaviour. The devices employed differed at ages 7 and 14, both in terms of the wear position (hip vs wrist) and the brand. Absolute differences are, however, likely to be small,27 and our different approaches to categorising participants at each time point based on distribution of raw acceleration versus thresholds to define MVPA volume produced a similar pattern of results. Our data showing that PA was relatively unstable over time are consistent with previous work on tracking of PA from childhood to adolescence.10 A lack of information regarding variations in PA and depressive symptoms in the intervening years between ages 7 and 14 is a limitation as depressive symptoms such as anhedonia, psychomotor retardation and fatigue during the interval between baseline and follow-up may have contributed to lower PA levels.
Comparison with existing studies
Previous observational studies have been unable to robustly assess PA–depression associations due to several methodological constraints. As indicated, there is a paucity of prospective studies, in particular none that has captured PA change at separate time points. The only two longitudinal studies to have investigated associations of device-measured PA with depression produced conflicting findings28 29; MVPA assessed at baseline was associated with lower risk of major depression in 6-year-old children followed up over a 4-year period,28 although no such association was found in a cohort of teenagers followed for approximately 3 years.29 The impact of LPA was not examined in either study. Previous work using self-reported PA has largely shown positive associations with MVPA.8 Self-reported MVPA is often derived from questions on sports participation, which may have favourable effects on mental health through, for example, social support mechanisms. In contrast, device-measured MVPA is devoid of any context and simply reflects movement, possibly explaining the mixed findings. In adult populations some data suggest that light/moderate intensity PA has greater antidepressive effects,30 benefits on positive mood31 and on reducing symptoms of fatigue compared with vigorous intensity.32
Consistent with previous work33 the prevalence of depressive symptoms was higher in girls in our study. Associations between PA and depressive symptoms were also far stronger in girls. The reasons for these sex differences are unclear. In boys the type and intensity of PA may be more important (eg, team sports). Biological mechanisms might explain higher vulnerability to depression in adolescent girls.34 Thus, PA may help partly stabilise biological imbalances during adolescence linked to depression. Social mechanisms may also exist; for example, body dissatisfaction and self-esteem may be reasons for inactivity in girls which may partly drive mental health.
Meaning of the study: possible explanations and implications for clinicians and policymakers
Treatment considerations for clinical depression in young people suggest that patients should be offered advice on the benefits of regular PA.35 Given that even subclinical symptoms in adolescence are associated with adverse outcomes, it is suggested that attention should also focus on PA in the prevention of depression, promoting intervention far earlier.
Conclusion
Using observational data to examine within-individual variation over a 7-year follow-up period, we demonstrate that increases in PA are associated with a lower risk of depressive symptoms in girls. Policies to increase PA might benefit from focusing on mid-childhood to early adolescence, when PA is particularly changeable.
What is already known on this subject
Observational studies suggest that higher levels of physical activity (PA) are associated with a lower risk of depressive symptoms in younger people.
Evidence is hampered by a lack of prospective data, a reliance on self-reported measures of activity and concerns regarding the time-varying characteristics of PA.
What this study adds
We use longitudinal data with repeat assessments of device-measured PA to examine prospective associations with depressive symptoms.
Girls who transitioned from low baseline PA to higher PA at follow-up had lower odds of depressive symptoms.
Participants who consistently met the PA guidelines of 60 min/day moderate to vigorous intensity at both 7 and 14 years of age experienced the lowest risk of depressive symptoms.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Correction notice This article has been corrected since it first published online. In the opening line of the Abstract ‘weak evidence’ has been replaced by ‘methodological shortcomings.'
Contributors MH (guarantor) had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. MH obtained funding, conceptualised and designed the study, performed the analyses, drafted the initial manuscript, and approved the final manuscript as submitted. PP, SB and DB conceptualised and designed the study, provided statistical input and critical revision of the manuscript, and approved the final manuscript as submitted.
Funding This research was conducted under the auspices of the Cross-Cohort Research Programme and was funded by the Economic and Social Research Council (grant number ES/M008584/1). GDB is supported by the UK Medical Research Council (MR/P023444/1) and the US National Institute on Aging (1R56AG052519-01; 1R01AG052519-01A1). The funders had no role in the study design; in the collection, analysis and interpretation of data; in writing of the report; or in the decision to submit the paper for publication.
Disclaimer The lead author (guarantor) affirms that the manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval Ethical approval was granted by the Northern and Yorkshire Multi-Centre Research Ethics Committee of the NHS.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data are available in a public, open access repository; https://www.ukdataservice.ac.uk/.