Contents lists available at ScienceDirect
Journal of Affective Disorders
journal homepage: www.elsevier.com/locate/jadPhysical activity in early adolescence predicts depressive symptoms 3 years
later: A community-based study
Johan Isaksson
a,b,⁎, Eva Noren Selinus
c,d,e, Cecilia Åslund
c,f, Kent W Nilsson
c,g a Department of Neuroscience, Child and Adolescent Psychiatry Unit, Uppsala University, Uppsala, Swedenb Department of Women's and Children's Health, Pediatric Neuropsychiatry Unit, Centre for Neurodevelopmental Disorders at Karolinska Institutet (KIND), Karolinska Institutet, Stockholm, Sweden
c Centre for Clinical Research, Västmanland County Hospital Västerås, Uppsala University, Sweden d The Swedish School of Sport and Health Sciences, Stockholm, Sweden
e Department of Clinical Neuroscience, Centre for Psychiatry Research & Education, Karolinska Institutet & Stockholm County Council, Stockholm, Sweden f Department of Public Health and Caring Sciences, Uppsala University, Sweden
g The School of Health, Care and Social Welfare, Mälardalen University, Sweden
A R T I C L E I N F O Keywords: Depression Sex differences Anxiety Physical activity ADHD Conduct problems A B S T R A C T
Background: Physical activity (PA) during adolescence is associated with a wide range of health benefits,
in-cluding lower levels of internalizing and externalizing problems. Although the association between PA and mental health has been established, there are few prospective studies investigating if the association between PA and internalizing/externalizing symptoms remains after adjustment for the baseline occurrence of such symp-toms, and those exploring any sex-specific pattern of the association.
Methods: Swedish adolescents (N = 1428; mean age = 14.38 years) were assessed and followed up 3 years later.
Self-reported data were collected for PA (recoded as low, moderate and high levels), internalizing (depression and anxiety) and externalizing (attention-deficit/hyperactivity disorder and disruptive behaviours) symptoms. A full path analysis was used to determine the main and interaction effects of PA and sex on internalizing/ex-ternalizing symptoms 3 years later, adjusting for these symptoms at baseline.
Results: Higher levels of PA were correlated with lower internalizing/externalizing symptoms. In the full path
analysis, PA during early adolescence predicted lower levels of depressive symptoms, but not anxiety or ex-ternalizing problems, 3 years later. A sex-specific effect of PA on depressive symptoms was found, wherein boys, but not girls, with high levels of PA showed reduced symptoms.
Limitations: Including parental ratings, diagnostic assessments and objective measures of PA would have
pro-vided additional information to the study.
Conclusions: Low levels of PA during early adolescence are a unique predictor for the development of depressive
symptoms among boys. PA should be considered when discussing prevention and treatment for depression in adolescents.
1. Introduction
Physical activity (PA) during childhood and adolescence has been associated with a wide range of health benefits, including improvement in indicators of physical health such as blood cholesterol level, blood pressure and obesity (Janssen and LeBlanc, 2010), as well as indicators for mental health (Biddle and Asare, 2011; Biddle et al., 2019). Mental health problems in children and youth are often classified into ex-ternalizing and inex-ternalizing symptoms (McElroy et al., 2018). Ex-ternalizing symptoms are characterized by behavioural disinhibition
and include conduct problems, inattention and hyperactivity, whereas internalizing symptoms refer to negative mood states and inhibition and include symptoms of depression and anxiety (McElroy et al., 2018). Even though most research on the association between PA and mental health has been conducted on adult populations, recent reviews on both clinical and community samples have stressed that PA inter-vention for children and adolescents reduce levels of internalizing and externalizing symptoms in general (Spruit et al., 2016), and for symp-toms of depression (Biddle et al., 2019) and attention-deficit/hyper-activity disorder (ADHD) (Ng et al., 2017), with effect sizes ranging
https://doi.org/10.1016/j.jad.2020.09.008
Received 20 February 2020; Received in revised form 25 August 2020; Accepted 2 September 2020
⁎Corresponding author.
E-mail address: johan.isaksson@neuro.uu.se (J. Isaksson).
Available online 07 September 2020
0165-0327/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
from small to moderate. In addition, reviews on prospective observa-tional studies have found PA to be negatively associated with sub-sequent symptoms of depression (Mammen and Faulkner, 2013; Schuch et al., 2018) and anxiety (Schuch et al., 2019).
The hypothesized positive effect of PA on mental health outcomes has been attributed to a variety of factors. These factors include phy-sical changes such as the release of endorphins and neurotransmitters, elevated blood and oxygen flow to the brain, heightened nerve cell growth and brain plasticity, decreased inflammation and blood pres-sure, and higher heart rate variability (Cotman and Berchtold, 2002; Ide and Secher, 2000; May et al., 2017; Mikkelsen et al., 2017). Moreover, PA has been suggested to offer learning opportunities within the social context of which the activity is performed, including the development of social competences, increased ability to resolve con-flicts and to collaborate with peers (Spruit et al., 2016). However, the association between PA and mental health may also be explained by low levels of PA among individuals with internalizing and externalizing problems. To examine any temporal sequence between PA and mental health problems, prospective studies are needed (Mammen and Faulkner, 2013).
Even though much research has been conducted on PA and mental health, the association between increased PA and reduced internalizing and externalizing symptoms are most evident in interventional studies, whereas observational studies have yielded more mixed results, with no clear support for whether PA precedes internalizing or externalizing symptoms and with a paucity of prospective studies investigating any sex-specific pattern of the association or any dose–response relationship (for a more comprehensive review see Biddle et al., 2019). Further-more, given the high overlap between externalizing and internalizing symptoms (McElroy et al., 2018), as well as the unclear temporal se-quence between PA and mental health problems (Mammen and Faulkner, 2013), there is a need to adjust for both of these transdiag-nostic clusters at baseline to evaluate any unique impact of PA on future symptoms. Even though, for instance, studies in general populations have shown that a low preference for physically active play at 8 years of age is associated with higher scores on inattention symptoms at 16 years of age (Khalife et al., 2014) and a greater self-rated energy ex-penditure at 16–17 years of age predicted reduced ADHD symptoms 3 years later in a sample of monozygotic twins while adjusting for un-measured confounding factors (Rommel et al., 2015), the impact of coexisting symptoms has not been taken into consideration.
To address this issue, we utilized a prospective research programme design and followed up a large cohort of Swedish adolescents. The aim of the present study was to evaluate any effects of self-reported PA on symptoms of depression, anxiety, ADHD and conduct problems after 3 years, adjusting for the baseline occurrence of these internalizing and externalizing symptoms to elucidate any unique effect of PA. In addi-tion, because there is evidence that sex differences may exist with re-gard to all types of mental health problems (Rescorla et al., 2012), we also investigated any sex-specific effect on the association between PA and symptoms of depression, anxiety, ADHD and conduct problems. 2. Methods
2.1. Participants and procedure
The data for this study were obtained from a large-scale prospective study on risk and protective factors for emotional and behavioural problems in Sweden, the SALVe cohort (Vadlin et al., 2018). All ado-lescents born in the Swedish county of Västmanland in 1997 and 1999 (and their parents) were contacted by regular mail and invited to par-ticipate in this study by completing a self-report questionnaire. The response rate was 40%. Västmanland is a medium-sized county in Sweden and considered reasonably representative of Swedish society in regards to its allocation of urban and rural areas, educational, income, employment and immigration levels (Sweden, 2014). After excluding
participants who lacked data on any variable (n = 554), the study sample comprised of 1428 adolescents assessed at wave 1 (58.1% girls; mean age = 14.38 years, standard deviation = 1.04) and followed up 3 years later at wave 2 (mean age = 17.31 years, standard devia-tion = 1.04). Those excluded had higher ratings of depressive symp-toms at wave 1 (t = 2.59; p = 0.10) and wave 2 (t = 3.60; p < 0.001), ADHD symptoms at wave 2 (t = 2.62; p = 0.09), conduct problems at wave 2 (t = 2.48; p = 0.13), were older (t = 1.98; p = 0.48), more frequently boys (χ² = 19.73; p < .001) and of non-Nordic ethnicity (χ² = 4.87; p = .027). In the study sample, most participants were of Scandinavian origin (80.1%), had parents who lived together (71.0%) and had an occupation or were students. This study was approved by the Ethical Review Board in Uppsala (Approval number: Dnr 2012/ 187). All of the adolescents and their parents provided written informed consent to participate in the study.
2.2. Measures
PA was measured using one item: ‘How often do you exercise at least 30 minutes in your spare time so that you get tired/sweaty?’ The adolescents reported on a seven-point scale from never (1) to every day (7). To assess the degree of PA, the participants’ responses were cate-gorized into low (never to 1–3 times a month [0]), moderate (1–3 times a week [1]) and high (≥4 times a week [2]) levels of PA.
As a proxy for socioeconomic status (SES), we used the parents’ level of education (as rated by the parents), coded as 1 = elementary school, 2 = secondary school and 3 = university, with a possible range of 2–6 for mother and father combined, with higher scores indicating higher SES.
Symptoms of depression were measured at both time points using the Depression Self-Rating Scale (DSRS; Svanborg and Ekselius, 2003). The DSRS has been adapted to be used for adolescents and to cover the A criterion for depression in the Diagnostic and Statistical Manual of Mental Disorders (DSM) IV. The DSRS has shown good psychometric properties and has been suggested to be used as an independent tool in epidemiological studies to provide a rough estimate of major depressive disorders in different population settings (Svanborg and Ekselius, 2003). The adolescents answered 15 questions about the presence of depressive symptoms during the last 2 weeks, including if they felt dysphoric/had an irritable mood, loss of interest/pleasure, sleep disturbances, fatigue, or loss of energy, feelings of worthlessness or guilt. Response options was ‘yes’ or ‘no’. A depression index was created by summation of scores of the reported symptoms. Because some items on the DSRS refer to the same symptom, we only counted each symptom once, resulting in a possible range of 0–9. The value of Cronbach's α was 0.81 at wave 1 and 0.77 after 3 years.
Symptoms of anxiety were assessed using the Spence Children's Anxiety Scale (Spence, 1997), which comprises 38 questions to which the adolescents respond on a 4-point scale (0 = never, 1 = sometimes, 2 = often and 3 = always). The scale has shown good to excellent internal reliability, and concurrent and discriminant validity, in both population-based (Essau et al., 2011) and clinical populations (Olofsdotter et al., 2016). The total score has a possible range of 0–114, where higher scores indicate more anxiety symptoms. The value of Cronbach's α was 0.87 at wave 1 and 0.89 after 3 years.
ADHD symptoms were assessed using the short version of the Adult ADHD Self-Report Scale, Adolescent version (ASRS-A-S), enquiring about symptoms of inattention and hyperactivity the last 6 months, using a six-item self-rating scale with response options ranging from never = 0 to very often = 4 (Kessler et al., 2005). The items correspond to the diagnostic criteria for ADHD and the scale has shown high in-ternal consistency and concurrent validity in both general and clinical populations (Kessler et al., 2005; Sonnby et al., 2015). The total score has a possible range of 0–24, where higher scores indicate more ADHD symptoms. The value of Cronbach's α for the total scale was 0.78 at wave 1 and 0.77 after 3 years.
Conduct problems were measured using a 26-item questionnaire assessing common disruptive behaviours, including norm-breaking be-haviour, vandalism, violence and stealing (Åslund et al., 2011; Nilsson et al., 2014; Isaksson et al., 2020). All of the items reflect the criteria of the DSM-IV diagnosis of conduct disorder. Adolescents in-dicated the frequency of conduct problems on a five-point scale (0 = never, 1 = once, 2 = 2–4 times, 3 = 5–10 times and 4 = >10 times), with a possible range of 0–104, with higher scores indicating more conduct problems. The value of Cronbach's α for the total scale was 0.80 at wave 1 and 0.85 after 3 years.
2.3. Statistical analyses
The Statistical Package for the Social Sciences (SPSS v25, IBM, Armonk, NY, USA) was used to analyse correlations between the vari-ables (Spearman's rho between ordinary and continuous varivari-ables, and the point-biserial correlation coefficient between a binary and con-tinuous variable), sex differences in symptoms and PA with t-test and χ², and difference in internalizing/externalizing symptoms by level of PA (low, moderate and high) with one-way analysis of variance (ANOVA) and Tukey post-hoc test. Since no multicollinearity was found (the variance inflation factor was below 2.0) we included all variables in the same model. For the main analysis a full path analysis (R Version 3.5.1; lavaan package) using robust maximum likelihood estimation was performed to assess the association between PA at wave 1 and symptoms of depression, anxiety, ADHD and conduct problem at wave 2 three years later, also adjusting for the same symptoms at wave 1. We also included age, sex, SES, and ethnicity in the path analysis, adjusting for the intercorrelation of all independent and dependent variables (see Supplementary Figure 1). As a second step, we included an interaction term (sex × PA) in the adjusted model. We present the standardized (regression) coefficients from the path analysis, where β refer to the change in standard deviations of the dependent variable per standard deviation increase in the predictor variable, and the R-square refers to the proportion of the variance in the dependent variable that is pre-dictable from the independent variables. Two-tailed tests with p-values <0.05 were considered as being statistically significant.
3. Results
Among the adolescents, 17.6% rated low levels, 45.5% rated mod-erate levels and 36.9% rated high levels of PA. As shown in Table 1, PA was negatively correlated with symptoms of depression, anxiety, ADHD and conduct problems. There was a sex difference in PA (χ² = 10.52; p = .005) with girls reporting more of moderate levels of PA (48.4% of the girls compared to 41.6% of the boys) and boys reporting more of high levels of PA (41.7% of the boys compared to 33.4% of the girls). Girls had higher scores on depressive symptoms (t = 7.32; p < 0.001, at wave 1; t = 12.62; p < 0.001, wave 2), anxiety (t = 13.29; p < 0.001, wave 1; t = 17.57; p < 0.001, wave 2) and ADHD (t = 5.78; p < 0.001, at wave 2), whereas boys had higher ratings on conduct problems (t = 6.05; p < 0.001, wave 1; t = 6.24; p < 0.001, wave 2). As shown in Table 2, those with high levels of PA during wave 1 had lower concurrent and subsequent ratings on symptoms of depression, anxiety, ADHD and conduct problems compared to those with low le-vels of PA, whereas those with moderate PA had lower rating compared to peers with low levels of PA on symptoms of depression, ADHD and conduct problems only at wave 1.
The full path analytic model is presented in Table 3. In the full model, 25% and 42% of the variance in symptoms of depression and anxiety was respectively explained. The comparable figures for the ADHD symptoms and conduct problems were 25% and 41%, respec-tively. Higher levels of PA were associated with lower scores on de-pressive symptoms, but not ADHD symptoms or conduct problems, while there was a trend for lower anxiety (p = 0.051). All symptoms at baseline predicted the same symptoms 3 years later. In addition, female
sex was associated with higher ratings on symptoms of depression, anxiety and ADHD symptoms, whereas male sex was associated with more conduct problems in the full model. Increasing age was associated with lower ratings on anxiety. An interaction effect for PA by sex on depressive symptoms was found (β = 0.19; p = 0.026), whereby boys displayed a larger decrease than girls in depressive symptoms (β = - 0.22 [CI = -0.37, -0.07] for boys and β = -0.04 [-0.16, 0.08] for girls). More specifically, for boys the DSRS ratings were .80 points lower for each increase in PA and the corresponding decrease for girls were .43. The association between PA and depressive symptoms was only sig-nificant for boys (p = 0.004).
4. Discussion
In this prospective study, we examined the association between self- reported PA and concurrent and subsequent psychiatric symptoms 3 years later in a community-based cohort of adolescents. Even though high levels of PA were associated with fewer concurrent and subsequent internalizing and externalizing symptoms, only the association between higher levels of PA during early adolescence and fewer depressive symptoms 3 years later remained when adjusting for all symptoms at baseline. A sex-specific effect of PA on depressive symptoms was found, where boys who reported greater PA showed a lower level of depressive symptoms than girls 3 years later.
Our results corroborate previous findings on the association be-tween PA and internalizing symptoms in adolescence (Biddle et al., 2019; Schuch et al., 2018; Spruit et al., 2016), and add to previous research by demonstrating that PA may be a unique predictor of symptoms of depression 3 years later, even when adjusting for these symptoms at baseline. In the review by Biddle et al. (2019) on PA and mental health in children and adolescents, it was concluded that there is no clear support for a dose–response relationship in the previous literature and that more research is needed on this subject. However, in our study, we found a dose–response associated relationship between high levels of PA (defined as exercising at least four times a week), and lower levels of depressive symptoms at both time points in the ANOVA, compared to low and medium levels of PA. Also, increasing levels of self-reported PA predicted fewer depressive symptoms in the full path analysis 3 years later. In line with this finding, the World Health Or-ganization recommends 60-minute daily PA of moderate to vigorous intensity, including vigorous-intensity activities at least 3 times per week, for children and youth (World Health Organization, 2010).
While PA was associated with reduced symptoms of depression three years later in our model, consideration needs to be taken to the fact that low physical activity is inherent in the criteria for depression, e.g., a reduction of physical movement, fatigue or loss of energy (American Psychiatric Association, 2013). It is therefore not surprising that low PA predict depressive symptoms. Nevertheless, since we ad-justed for depressive symptoms at wave 1 in the path analysis, the re-sults indicate that PA may have a unique effect on depressive symp-toms. The positive effect of PA on internalizing symptoms has been attributed to a variety of factors and several hypotheses have been suggested to mediate this effect (Mikkelsen et al., 2017). These theories include an increase in levels of endorphins; an increase in body tem-perature resulting in a decrease in muscular tension and feelings of overall relaxation; an increase in neurotransmitters affecting ser-otonergic and adrenergic levels in the brain, which may act in the same way as the serotonin reuptake inhibitor antidepressants; and an at-tenuation in the cortisol response via the hypothalamic–pituitary–a-drenal axis to stress (Mikkelsen et al., 2017). In addition, the effects may also be attributed to psychosocial factors such as increased feelings of self-efficacy and self-regulation (Mikkelsen et al., 2017).
We found that the association between PA and depressive symptoms was sex-specific, with a larger reduction in depressive symptoms 3 years later for boys compared to girls when reporting higher levels of PA at baseline. In addition, the association was only significant for
boys. The role of possible moderators such as sex, which may influence the relationship between PA and depression, have not been thoroughly explored and findings from previous research are inconclusive. In line with our finding, in a study conducted on a large sample of Norwegian young adolescents, sedentary activities predicted depressive symptoms 1 year later among boys but not girls (Sund et al., 2011). In contrast, Mammen and Faulkner (2013) found that a protective effect of PA on depression was more specific to girls than boys which suggests that girls may benefit more from the social aspects of PA than males. In a recent meta-analysis summarizing prospective cohort studies conducted in both child- and adult populations, the potential protective association of PA was found to be similar for both males and females (Schuch et al., 2018). An important difference between our study and previous re-search is the inclusion of a wider range of transdiagnostic symptoms, and hypothetically, other comorbidities such as externalizing problems and anxiety may blur the association between PA and depressive symptoms in girls. To conclude, more research is needed to assess the effect of sex on the association between PA and internalizing symptoms. In line with previous research, we found that girls had higher rat-ings on symptoms of depression and anxiety whereas boys had higher ratings on conduct problems (Liu et al., 2011; Noren Selinus et al., 2016; Rescorla et al., 2012), a sex difference that seems to be more evident among adolescents than among younger children (Costello et al., 2002; Ruchkin & Schwab-Stone, 2003). Interestingly, girls rated more ADHD symptoms at wave 2 compared to boys, a finding that contrasts previous reports of ADHD being more common among boys (Franke et al., 2018). However, the often-reported sex difference in ADHD is more pronounced in clinical samples. In com-munity-based studies, using self-ratings, the prevalence of ADHD symptoms seems to be more equally distributed between girls and boys (Sonnby et al. 2011, 2015). Regarding sex differences in PA, our finding was more mixed, with girls reporting more moderate levels and boys’ high levels of PA. This partly corroborates with previous research with boys reporting more PA across countries (de Looze et al., 2019).
PA did not predict externalizing symptoms 3 years later in our study. This finding is in contrast with those of a recent review by Ng et al. (2017), where it was concluded that both short-term and long- term PA interventions result in clinical benefits for individuals with ADHD. In addition, in another review and meta-analysis, including randomized controlled trial studies of intervention programmes with aerobic and yoga exercise, evidence indicated that short-term aerobic exercise appears to be effective for mitigating symptoms such as at-tention deficit, hyperactivity and impulsivity and related symptoms like anxiety, executive function and social disorders in children with ADHD (Cerrillo-Urbina et al., 2015). However, in contrast to our study, co-morbidities were not fully adjusted for in these reviews, which may explain our contradictory finding. PA was also associated with fewer symptoms of ADHD 3 years later in our unadjusted models, i.e., in the correlation analysis and in the ANOVA. Another possible explanation for our contradictory finding may reflect the diagnostic criteria of ADHD as a neurodevelopmental disorder present at childhood. ADHD symptoms do not vary over time to the same extent as symptoms of depression and anxiety, limiting the range of possible symptom change. Our study has several limitations. First, mental health problems are influenced by many factors, including genetic and environmental fac-tors (Segurado et al., 2011; Visscher et al., 2012), which were not in-cluded in this study. Second, the data were based on self-ratings and neither parental ratings nor diagnostic assessments were used. We also lack measurements on degree of functional impairment. Third, our measure on PA has not been used in previous research, however, the use of categorical response options, similar to our study, have been shown to provide superior validity to open answers when asking for level of PA (Olsson et al., 2016). The use of objective measures of PA could have provided additional information; moreover, previous studies using an ActiGraph accelerometer to measure PA have not confirmed the association between PA and internalizing symptoms (Opdal et al.,
Table 1 Correlations (Spearman's rho for ordinal and continuous data, and the point-biserial correlation coefficient between a binary 1 and continuous/ordinal variable) between the study variables. PA Depressivesymptoms w1 Anxietysymptoms w1 ADHDsymptoms w1 Conduct problems w1 Depressivesymptoms w2 Anxietysymptoms w2 ADHDsymptoms w2 Conduct problems w2 Sex(female) Age SES Depressive symptoms w1 –.17 ⁎⁎⁎ 1.00 Anxiety Symptoms w1 –.13 ⁎⁎⁎ .54 ⁎⁎⁎ 1.00 ADHD Symptoms w1 –.13 ⁎⁎⁎ .49 ⁎⁎⁎ .45 ⁎⁎⁎ 1.00 Conduct problems w1 –.06 * .32 ⁎⁎⁎ .21 ⁎⁎⁎ .42 ⁎⁎⁎ 1.00 Depressive symptoms w2 –.15 ⁎⁎⁎ .37 ⁎⁎⁎ .40 ⁎⁎⁎ .27 ⁎⁎⁎ .13 ⁎⁎⁎ 1.00 Anxiety Symptoms w2 –.14 ⁎⁎⁎ .33 ⁎⁎⁎ .61 ⁎⁎⁎ .27 ⁎⁎⁎ .06 * .63 ⁎⁎⁎ 1.00 ADHD Symptoms w2 –.11 ⁎⁎⁎ .30 ⁎⁎⁎ .32 ⁎⁎⁎ .46 ⁎⁎⁎ .24 ⁎⁎⁎ .50 ⁎⁎⁎ .48 ⁎⁎⁎ 1.00 Conduct problems w2 –.05 .20 ⁎⁎⁎ .11 ⁎⁎⁎ .30 ⁎⁎⁎ .56 ⁎⁎⁎ .23 ⁎⁎⁎ .13 ⁎⁎⁎ .34 ⁎⁎⁎ 1.00 Sex (girl) –.07 * .19 ⁎⁎⁎ .33 ⁎⁎⁎ .05 –.16 ⁎⁎⁎ .32 ⁎⁎⁎ .42 ⁎⁎⁎ .15 ⁎⁎⁎ –.16 ⁎⁎⁎ 1.00 Age –.03 .12 ⁎⁎⁎ .09 ⁎⁎ .12 ⁎⁎⁎ .17 ⁎⁎⁎ .05 –.00 .03 .09 ⁎⁎⁎ .02 1.00 SES .18 ⁎⁎⁎ –.12 ⁎⁎⁎ –.05 –.10 ⁎⁎⁎ –.07 ⁎⁎ –.04 –.06 * –.05 * –.03 –.03 –.05 1.00 Ethnicity (non-Nordic) –.07 ⁎⁎ .08 ⁎⁎ .04 –.00 –.03 .08 ⁎⁎ .04 –.01 –.01 .00 .01 –.07 Note: 1All variables are continuous except for PA and SES which are ordinal, and sex and ethnicity which are binary. ADHD, attention-deficit/hyperactivity disorder; PA, physical activity; SES, socioeconomic status; w1, wave one; w2, wave two. ⁎p < .05. ⁎⁎ p < .01. ⁎⁎⁎ p < .001.
2019). Fourth, our study lacks data on any interventions targeting mental health problems that the participants may have received during the study time, which may have an impact on the outcomes. Fifth, our study population was largely homogeneous, with most participants of Nordic origin. A more heterogeneous sample may result in different or complementary findings. Sixth, the high rate of attrition that occurred over the course of the study period is common in longitudinal studies where individuals lost to follow-up also often differ on several char-acteristics from those retained (Saiepour et al., 2019). A high attrition may result in a selection bias that may hypothetically affect estimates of incidence, prevalence, and estimates of any association. However, since previous research has reported that loss to follow-up rarely affects es-timates of association (Saiepour et al., 2019; Wolke et al., 2009) and since methods for adjusting loss of cases, such as multiple imputation, involve assumptions that the data is missing at random, we choose a more stringent approach and calculated how the excluded participants differed from the included participants on the study variables.
We conclude that low levels of PA during early adolescence appear to be a unique predictor for the development of depressive symptoms 3 years later, and that this association is significant only for boys. Our findings suggest that PA may be of relevance for early prevention efforts and should be included when discussing treatment options for depres-sion in adolescents.
Authors' contribution
Author JI conducted the statistical analyses and drafted the manu-script in collaboration with ENS. Author CÅ and KWN designed the study and were responsible for the data collection. All authors read and approved the final manuscript.
Funding
This research was supported by grants from the Uppsala University Hospital Research Fund (ALF), the Swedish Brain Foundation (Hjärnfonden), the Swedish Alcohol Monopoly Research Council (SRA), the Swedish Council for Working Life and Social Research (FAS), the Uppsala and Örebro Regional Research Council, the Fredrik and Ingrid Thurings Foundation, the County Council of Västmanland, the Svenska Spel Research Foundation, the Söderström König Foundation (SLS- 559921, SLS-655791, SLS-745221), Åke Wiberg's Foundation (M15- 0239), and the Swedish Research Council for Health, Working Life and Welfare (FORTE) (2015-00897). The funding sources had no further role in study design; the data collection, analysis, and interpretation of data, the writing of the report; or the decision to submit the paper for publication.
Table 2
Group and post-hoc comparison for levels of physical activity (PA) on symptoms of depression, anxiety, ADHD and conduct problems wave 1 and wave 2 (Mean and SD).
Low (L) PA Moderate (M) PA High (H) PA ANOVA
Depressive symptoms Wave 1 Wave 2 2.51 (2.41) 3.26 (2.70) 1.81 (2.10) 3.03 (2.58) 1.46 (1.97) 2.32 (2.43) F=21.08*** (L>M***, L>H***, M>H*) F=15.79*** (L>H***, M>H***) Anxiety symptoms Wave 1 Wave 2 21.16 (14.18) 25.07 (15.99) 19.19 (13.08) 23.42 (15.81) 16.42 (11.90) 19.74 (14.03) F=13.15*** (L>H***, M>H**) F=14.41*** (L>H***, M>H***) ADHD symptoms Wave 1 Wave 2 8.27 (4.81) 9.82 (5.26) 6.90 (4.31) 9.49 (4.65) 6.39 (4.01) 8.61 (4.57) F=16.39*** (L>M***, L>H***) F=7.49** (L>H**, M>H**) Conduct problems Wave 1 Wave 2 6.37 (7.91) 7.15 (8.76) 5.09 (5.50) 6.60 (6.59) 4.61 (4.81) 5.76 (6.25) F=7.94*** (L>M**, L>H***) F=4.05* (L>H*) Table 3
Full path analytic model of symptoms of depression, anxiety, ADHD and conduct problems regressed on levels of PA and covariates, with standardized regression coefficients (β) and 95% confidence intervals (CI), also adjusting for the intercorrelation of independent and dependent variables (not shown in Table).
Parameter estimated Depressive symptoms W2 Anxiety W2 ADHD symptoms W2 Conduct problems W2
β CI β CI β CI β CI PAa -.06* -.11, -.01 -.04 -.08, .00 -.02 -.07, .03 -.01 -.05, .03 Sex (girl) .23⁎⁎⁎ .18, .28 .25⁎⁎⁎ .21, .30 .11⁎⁎⁎ .06, .16 -.07⁎⁎ -.11, -.02 Age .00 -.02, .05 -.05* -.09, -.01 -.04 -.08, .01 -.02 .02, -.06 SES .03 -.02, .07 -.03 -.07, .01 .01 -.03, .06 -.01 -.05, .03 Ethnicity (Non-Nordic) .06* .01, .11 .02 -.02, .06 -.01 -.06, .03 .01 -.04, .05 Depressive symptomsb W1 .18⁎⁎⁎ .12, .25 -.03 -.08, .03 .04 -.02, .10 -.03 -.08, .03 Anxietyc W1 .16⁎⁎⁎ .11, .23 .50⁎⁎⁎ .45, .56 .07* .01, .13 -.00 -.06, .05 ADHD symptomsd W1 .05 -.01, .02 .05* .00, .11 .39⁎⁎⁎ .33, .44 .08⁎⁎ .03, .12 Conduct problemse W1 .07* .01, .12 -.00 -.05, .04 .06* .01, .12 .60⁎⁎⁎ .55, 65 R-square .25⁎⁎⁎ .42⁎⁎⁎ .25⁎⁎⁎ .41⁎⁎⁎
ADHD, attention-deficit/hyperactivity disorder; PA, Physical activity; SES, socioeconomic status; W1, Wave 1; W2, Wave 2
a Low levels of PA (coded as 0), moderate levels (1), high levels (2). b Measured with the Depression Self-Rating Scale.
c Measured with the Spence Children's Anxiety Scale.
dMeasured with the Adult ADHD Self-Report Scale, Adolescent version.
e Measured with a 26-item questionnaire assessing common disruptive behaviours. ⁎ p < .05.
⁎⁎ p < .01. ⁎⁎⁎ p < .001.
Declaration of Competing Interest None.
Acknowledgements
We are grateful to all the adolescents and parents who participated in the study.
Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.jad.2020.09.008.
References
American Psychiatric Association, 2013. Diagnostic and Statistical Manual of Mental Disorders, fifth ed. American Psychiatric Publishing, Arlington, VA.
Biddle, S.J., Asare, M., 2011. Physical activity and mental health in children and ado-lescents: a review of reviews. Br. J. Sports Med. 45 (11), 886–895.
Biddle, S.J., Ciaccioni, S., Thomas, G., Vergeer, I., 2019. Physical activity and mental health in children and adolescents: an updated review of reviews and an analysis of causality. Psychol. Sport Exerc. 42, 146–155.
Cerrillo-Urbina, A.J., Garcia-Hermoso, A., Sanchez-Lopez, M., Pardo-Guijarro, M.J., Santos Gomez, J.L., Martinez-Vizcaino, V., 2015. The effects of physical exercise in children with attention deficit hyperactivity disorder: a systematic review and meta- analysis of randomized control trials. Child Care Health Dev. 41 (6), 779–788. Costello, E.J., Pine, D.S., Hammen, C., March, J.S., Plotsky, P.M., Weissman, M.M., ...,
Leckman, J.F., 2002. Development and natural history of mood disorders. Biol. Psychiatry 52 (6), 529–542.
Cotman, C.W., Berchtold, N.C., 2002. Exercise: a behavioral intervention to enhance brain health and plasticity. Trends Neurosci. 25 (6), 295–301.
de Looze, M., Elgar, F.J., Currie, C., Kolip, P., Stevens, G., 2019. Gender inequality and sex differences in physical fighting, physical activity, and injury among adolescents across 36 countries. J. Adolesc. Health 64 (5), 657–663.
Essau, C.A., Sasagawa, S., Anastassiou-Hadjicharalambous, X., Guzman, B.O., Ollendick, T.H., 2011. Psychometric properties of the Spence Child Anxiety Scale with adoles-cents from five European countries. J. Anxiety Disord. 25 (1), 19–27.
Ide, K., Secher, N.H., 2000. Cerebral blood flow and metabolism during exercise. Prog. Neurobiol. 61 (4), 397–414.
Isaksson, J., Vadlin, S., Olofsdotter, S., Aslund, C., Nilsson, K.W., 2020. Psychotic-like experiences during early adolescence predict symptoms of depression, anxiety, and conduct problems three years later: a community-based study. Schizophr. Res. 215, 190–196.
Janssen, I., Leblanc, A.G., 2010. Systematic review of the health benefits of physical activity and fitness in school-aged children and youth. Int. J. Behav. Nutr. Phys. Act. 7, 40.
Kessler, R.C., Adler, L., Ames, M., Demler, O., Faraone, S., Hiripi, E., Howes, M.J., Jin, R., Secnik, K., Spencer, T., Ustun, T.B., Walters, E.E., 2005. The World Health Organization Adult ADHD Self-Report Scale (ASRS): a short screening scale for use in the general population. Psychol. Med. 35 (2), 245–256.
Khalife, N., Kantomaa, M., Glover, V., Tammelin, T., Laitinen, J., Ebeling, H., Hurtig, T., Jarvelin, M.R., Rodriguez, A., 2014. Childhood attention-deficit/hyperactivity dis-order symptoms are risk factors for obesity and physical inactivity in adolescence. J. Am. Acad. Child Adolesc. Psychiatry 53 (4), 425–436.
Liu, J., Chen, X., Lewis, G., 2011. Childhood internalizing behaviour: analysis and im-plications. J. Psychiatr. Ment. Health Nurs. 18 (10), 884–894.
Mammen, G., Faulkner, G., 2013. Physical activity and the prevention of depression: a systematic review of prospective studies. Am. J. Prev. Med. 45 (5), 649–657. May, R., McBerty, V., Zaky, A., Gianotti, M., 2017. Vigorous physical activity predicts
higher heart rate variability among younger adults. J. Physiol. Anthropol. 36 (1), 24. McElroy, E., Shevlin, M., Murphy, J., McBride, O., 2018. Co-occurring internalizing and externalizing psychopathology in childhood and adolescence: a network approach. Eur. Child Adolesc. Psychiatry 27 (11), 1449–1457.
Mikkelsen, K., Stojanovska, L., Polenakovic, M., Bosevski, M., Apostolopoulos, V., 2017. Exercise and mental health. Maturitas 106, 48–56.
Ng, Q.X., Ho, C.Y.X., Chan, H.W., Yong, B.Z.J., Yeo, W.S., 2017. Managing childhood and adolescent attention-deficit/hyperactivity disorder (ADHD) with exercise: a sys-tematic review. Complement Ther. Med. 34, 123–128.
Nilsson, K.W., Comasco, E., Hodgins, S., Oreland, L., Åslund, C., 2014. Genotypes do not confer risk for delinquency but rather alter susceptibility to positive and negative environmental factors: gene-environment interactions of BDNF Val66Met, 5-HTTLPR,
and MAOA-uVNTR [corrected]. Int. J. Neuropsychopharmacol. 18 (5), pyu107. Noren Selinus, E., Molero, Y., Lichtenstein, P., Anckarsater, H., Lundstrom, S., Bottai, M.,
Hellner Gumpert, C., 2016. Subthreshold and threshold attention deficit hyper-activity disorder symptoms in childhood: psychosocial outcomes in adolescence in boys and girls. Acta Psychiatr. Scand. 134 (6), 533–545.
Olofsdotter, S., Sonnby, K., Vadlin, S., Furmark, T., Nilsson, K.W., 2016. Assessing ado-lescent anxiety in general psychiatric care: diagnostic accuracy of the Swedish self- report and parent versions of the Spence Children's Anxiety Scale. Assessment 23 (6), 744–757.
Olsson, S.J., Ekblom, O., Andersson, E., Borjesson, M., Kallings, L.V., 2016. Categorical answer modes provide superior validity to open answers when asking for level of physical activity: a cross-sectional study. Scand. J. Public Health 44 (1), 70–76. Opdal, I.M., Morseth, B., Handegard, B.H., Lillevoll, K., Ask, H., Nielsen, C.S., Horsch, A.,
Furberg, A.S., Rosenbaum, S., Rognmo, K., 2019. Change in physical activity is not associated with change in mental distress among adolescents: the Tromso study: fit futures. BMC Public Health 19 (1), 916.
Rescorla, L., Ivanova, M.Y., Achenbach, T.M., Begovac, I., Chahed, M., Drugli, M.B., Emerich, D.R., Fung, D.S., Haider, M., Hansson, K., Hewitt, N., Jaimes, S., Larsson, B., Maggiolini, A., Marković, J., Mitrović, D., Moreira, P., Oliveira, J.T., Olsson, M., Ooi, Y.P., Petot, D., Pisa, C., Pomalima, R., da Rocha, M.M., Rudan, V., Sekulić, S., Shahini, M., de Mattos Silvares, E.F., Szirovicza, L., Valverde, J., Vera, L.A., Villa, M.C., Viola, L., Woo, B.S., Zhang, E.Y., 2012. International epidemiology of child and adolescent psychopathology ii: integration and applications of dimensional findings from 44 societies. J. Am. Acad. Child Adolesc. Psychiatry 51 (12), 1273–1283.e8. Rommel, A.S., Lichtenstein, P., Rydell, M., Kuja-Halkola, R., Asherson, P., Kuntsi, J.,
Larsson, H., 2015. Is physical activity causally associated with symptoms of attention- deficit/hyperactivity disorder? J. Am. Acad. Child Adolesc. Psychiatry 54 (7), 565–570.
Ruchkin, V., Schwab-Stone, M., 2003. What can we learn from developmental studies of psychiatric disorders? Lancet 362 (9400), 1951–1952.
Saiepour, N., Najman, J.M., Ware, R., Baker, P., Clavarino, A.M., Williams, G.M., 2019. Does attrition affect estimates of association: a longitudinal study. J. Psychiatr. Res. 110, 127–142.
Schuch, F.B., Stubbs, B., Meyer, J., Heissel, A., Zech, P., Vancampfort, D., Rosenbaum, S., Deenik, J., Firth, J., Ward, P.B., Carvalho, A.F., Hiles, S.A., 2019. Physical activity protects from incident anxiety: a meta-analysis of prospective cohort studies. Depress. Anxiety 36 (9), 846–858.
Schuch, F.B., Vancampfort, D., Firth, J., Rosenbaum, S., Ward, P.B., Silva, E.S., Hallgren, M., Ponce De Leon, A., Dunn, A.L., Deslandes, A.C., Fleck, M.P., Carvalho, A.F., Stubbs, B., 2018. Physical activity and incident depression: a meta-analysis of pro-spective cohort studies. Am. J. Psychiatry 175 (7), 631–648.
Segurado, R., Bellgrove, M.A., Manconi, F., Gill, M., Hawi, Z., 2011. Epistasis between neurochemical gene polymorphisms and risk for ADHD. Eur. J. Hum. Genet. 19 (5), 577–582.
Sonnby, K., Aslund, C., Leppert, J., Nilsson, K.W., 2011. Symptoms of ADHD and de-pression in a large adolescent population: co-occurring symptoms and associations to experiences of sexual abuse. Nord. J. Psychiatry 65 (5), 315e322.
Sonnby, K., Skordas, K., Olofsdotter, S., Vadlin, S., Nilsson, K.W., Ramklint, M., 2015. Validation of the World Health Organization Adult ADHD Self-Report Scale for adolescents. Nord. J. Psychiatry 69 (3), 216–223.
Spence, S.H., 1997. Structure of anxiety symptoms among children: a confirmatory factor- analytic study. J. Abnorm. Psychol. 106 (2), 280–297.
Spruit, A., Assink, M., van Vugt, E., van der Put, C., Stams, G.J., 2016. The effects of physical activity interventions on psychosocial outcomes in adolescents: a meta- analytic review. Clin. Psychol. Rev. 45, 56–71.
Sund, A.M., Larsson, B., Wichstrom, L., 2011. Role of physical and sedentary activities in the development of depressive symptoms in early adolescence. Soc. Psychiatry Psychiatr. Epidemiol. 46 (5), 431–441.
Svanborg, P., Ekselius, L., 2003. Self-assessment of DSM-IV criteria for major depression in psychiatric out- and inpatients. Nord. J. Psychiatry 57 (4), 291–296.
Sweden, S., 2014. Statistisk årsbok för Sverige 2014 [Statistical yearbook of Sweden 2014]. Statistics Sweden, Information and Publishing Department, Stockholm. Vadlin, S., Åslund, C., Nilsson, K.W., 2018. A longitudinal study of the individual- and
group-level problematic gaming and associations with problem gambling among Swedish adolescents. Brain Behav. 8 (4), e00949.
Visscher, P.M., Brown, M.A., McCarthy, M.I., Yang, J., 2012. Five years of GWAS dis-covery. Am. J. Hum. Genet. 90 (1), 7–24.
Wolke, D., Waylen, A., Samara, M., Steer, C., Goodman, R., Ford, T., Lamberts, K., 2009. Selective drop-out in longitudinal studies and non-biased prediction of behaviour disorders. Br. J. Psychiatry 195 (3), 249–256.
World Health Organization, 2010. https://www.who.int/dietphysicalactivity/factsheet_ young_people/en/(accessed 19 December 2019).
Åslund, C., Nordquist, N., Comasco, E., Leppert, J., Oreland, L., Nilsson, K.W., 2011. Maltreatment, MAOA, and delinquency: sex differences in gene-environment inter-action in a large population-based cohort of adolescents. Behav. Genet. 41 (2), 262–272.
