Journal of Adolescence 40 (2015) 83e92

Contents lists available at ScienceDirect

Journal of Adolescence journal homepage: www.elsevier.com/locate/jado

The effect of physical activity on depression in adolescence and emerging adulthood: A growth-curve analysis* Meghan L. McPhie 1, Jennine S. Rawana* Department of Psychology, York University, Toronto, ON M3J 1P3, Canada

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 23 February 2015

This study examined the influence of physical activity on the trajectory of depression from adolescence through emerging adulthood (EA). Using data from the National Longitudinal Study of Adolescent Health (Waves I to IV), latent growth curve modeling was performed to assess how physical activity and gender influenced depression across adolescence and EA. Higher levels of physical activity in mid-adolescence were associated with lower levels of depression during mid-adolescence and slower inclines and declines in depression over time. Boys had lower levels of depression in mid-adolescence and slower inclines and declines in depression over time compared to girls. Findings provide evidence that current theories on understanding depression and mental health prevention programs may be enhanced by the inclusion of physical activity. © 2015 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.

Keywords: Depression Physical activity Adolescence Emerging adulthood Latent growth curve modeling

Depression is a mental health issue facing individuals across the life span and is commonly associated with negative health, economic, and quality of life consequences. Depression typically emerges during adolescence (Kessler, Avenevoli, & Merikangas, 2001; Paus, Keshavan, & Giedd, 2008) and endures into adulthood (Rudolph, 2009). Noticeable gender differences exist with women and adolescent girls experiencing higher rates of depression than men and adolescent boys (Hyde, Mezulis, & Abramson, 2008), although these gender differences are not consistently found until mid-adolescence (Hankin et al., 1998; Nolen-Hoeksema & Girgus, 1994). Traditionally, research has focused on established risk factors for depression, such as emotion dysregulation (Joormann & D'Avanzato, 2010; Yap et al., 2011), and poor social support (Needham, 2008; Vaughan, Foshee, & Ennett, 2010). More recently, the literature focuses on factors that lessen one's likelihood of experiencing low mood. Physical activity is a behavioral strategy that has been linked to improved mood (Birkeland, Torsheim, & Wold, 2009; Jerstad, Boutelle, Ness, & Stice, 2010; McPhie & Rawana, 2012; Motl, Birnbaum, Kubik, & Dishman, 2004). However, the long-term influence of engagement in physical activity on depressive symptoms beyond adolescence and into the next developmental stage of life, namely emerging adulthood (EA; 18e29 age range), is relatively unknown. Thus, the current study explored the influence of

* This research uses data from Add Health, a program project directed by Kathleen Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Information on how to obtain the Add Health data files is available on the Add Health website (http://www.cpc.unc.edu/addhealth). No direct support was received from grant P01-HD31921 for this analysis. * Corresponding author. Tel.: þ1 (416) 736 2100x20771; fax: þ1 (416) 736 5184. E-mail addresses: [email protected] (M.L. McPhie), [email protected] (J.S. Rawana). 1 Tel.: þ1 (416) 736 2100x33456; fax: þ1 (416) 736 5184.

http://dx.doi.org/10.1016/j.adolescence.2015.01.008 0140-1971/© 2015 The Foundation for Professionals in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.

84

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

engagement in physical activity and the predictive effect of gender on the longitudinal trajectory of depressive symptoms from adolescence to EA using a national, longitudinal sample. Trajectories of depression across development The developmental psychopathology perspective suggests that depression is best conceptualized as a heterogeneous condition that occurs through a diversity of developmental pathways (Cicchetti & Toth, 1998). This perspective highlights the importance of examining developmental trajectories of psychopathology, such as depression, and using longitudinal research designs that allow one to examine the relationship between various risk and protective factors and long-term outcomes. As it currently stands, the literature on depression trajectories is limited in several respects. For one, relatively few studies have extended these findings further into EA, a period marked by significant changes and exploration, and important decisions regarding love and work (Arnett, 2000). Second, the literature has focused on the influence of risk and vulnerability factors. Third, few studies have examined gender differences in trajectories of depression that extend from adolescence through late EA, as the majority of research has tended to focus on either the adolescent or EA period independently, or a brief transitional period between these two stages of development. Thus, trajectory analyses represent a relatively new approach in which to examine depression pathways from adolescence to late EA. Physical activity The relationship between mental well-being and physical activity has attracted considerable attention in the literature. However, there are presently few studies on the interplay between physical activity and depressive symptoms among adolescents, as the majority of the research has focused on adult populations. Additionally, the paucity of studies using adolescent samples have generated mixed findings. Cross-sectional studies have consistently reported a negative relationship between physical activity and depressive symptoms; however, a temporal relationship cannot be assumed (Brand et al., 2010; Elliot, Kennedy, Morgan, Anderson, & Morris, 2012; Jerstad et al., 2010; McPhie & Rawana, 2012; Sigfusdottir, Asgeirsdottir, Sigurdsson, & Gudjonsson, 2011). Further, recent research has provided some encouraging support for physical activity as a beneficial intervention for mood in both clinical (Josefsson, Lindwall, & Archer, 2014; Mota-Pereira et al., 2011; Stanton, Happell, Hayman, & Reaburn, 2014) and non-clinical (Kalak et al., 2012) samples of adolescents and adults. Research has demonstrated that physical activity and depressed mood tend to negatively covary across time in both € rjesson, & Ahlborg, 2014); however, adolescent (Birkeland et al., 2009) and adult populations (Lindwall, Gerber, Jonsdottir, Bo few studies have examined the direction of the relationship between physical activity and depressive symptoms longitudinally during adolescence (Birkeland et al., 2009; Jerstad et al., 2010; Raudsepp & Neissaar, 2012). Birkeland et al. (2009) found no evidence for the influence of initial physical activity level on later changes in depressive symptoms among adolescents and early emerging adults. Moreover, Jerstad et al. (2010) found a bidirectional relation among physical activity and depression among adolescent girls, wherein physical activity was found to significantly reduce risk for future depressive symptoms and vice-versa. Raudsepp and Neissaar (2012) found that baseline level and changes in physical activity were negatively associated with baseline level and changes in depressed mood among adolescent girls. Thus, to date the literature has produced mixed findings, and the long-term effects of physical activity on depressive symptoms among male and female adolescents and emerging adults is unclear. The few studies that have examined the prospective connection between physical activity and depressive symptoms are limited in several ways. Studies are generally restricted to the adolescent period, have short follow-up periods (e.g., less than 10 years), and include mainly adolescent girls (e.g., Jerstad et al., 2010; Raudsepp & Neissaar, 2012). As such, the current study aimed to build on the existing literature by examining the connection between physical activity and depressive symptoms over a span of 14 years from mid-adolescence into late EA using a large, longitudinal sample of adolescent boys and girls. Current study The overarching objective of this study was to examine the predictive effect of engagement in physical activity and gender on the longitudinal trajectory of depressive symptoms from adolescence into EA. Individuals were followed from midadolescence (i.e., age 15) to late EA (i.e., age 28) to encompass the time period in which gender differences in depression are consistently observed (Nolen-Hoeksema & Girgus, 1994). As far as the authors are aware, this is the first study to predict a depression trajectory across adolescence and into later EA. The first study goal was to examine the prospective relationship between engagement in physical activity and depression. It was expected that engagement in physical activity would be associated with lower levels of depressive symptoms at baseline (age ¼ 15 years) and that initial levels of engagement in physical activity would be related to more gradual declines and inclines in depressive symptoms over development. The second objective was to examine gender differences regarding the trajectory of depression. It was expected that gender differences in depressive symptoms at baseline would exist, with adolescent girls having higher initial levels compared to adolescent boys. Moreover, it was expected that gender would significantly predict the rate of change in depressive symptoms across adolescence into EA in that boys would have steeper declines and more gradual inclines in depressive symptoms compared to girls.

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

85

A third objective was to test a gender-by-physical activity interaction in relation to the depressive symptom trajectory. No specific hypotheses were made given the exploratory nature of this research objective. Method Data and participants The present research involved secondary data analysis of a longitudinal, population-based sample from the restricted-use database of The National Longitudinal Study of Adolescent Health (Add Health; Harris, 2009; visit http://www.cpc.unc.edu/ projects/addhealth for more information on Add Health). The Institutional Review Board at the University of North Carolina at Chapel Hill provided ethical approval to obtain informed consent and carry out the Add Health study. Ethics approval was also obtained at the university to securely store the Add Health data. Every high school in the United States that had an 11th grade and at least 30 students enrolled in the school were eligible to be selected to partake in the study. A random sample of 80 high schools and 52 middle schools were chosen in relation to enrolment size and stratified by region of country, urbanicity, school size, school type, and ethnicity. More than 90,000 students in grades 7 to 12 participated in the in-school questionnaire and a sample of 27,000 students was selected to partake in the subsequent in-home interviews. The Add Health data used in the present study included multiple in-home questionnaires that followed participants over the course of four waves of data collection: 20,745 adolescents from Wave 1 (1995), 14,738 adolescents from Wave 2 (1996), 15,197 emerging adults from Wave 3 (2001/2002), and 15,701 emerging adults from Wave 4 (2008/2009). In order to capture the development through adolescence and into EA, all participants aged 15 at Wave 1 were included in this study (N ¼ 3676; 51% female). Therefore, adolescents who were either younger or older than 15 at Wave 1 were not included in the analyses. Thus, the mean age (in years) of participants at Wave 1, Wave 2, Wave 3, and Wave 4 was 15.0 (SD ¼ 0.00), 15.9 (SD ¼ 0.33), 21.4 (SD ¼ 0.50), 27.9 (SD ¼ 0.48), respectively. Measures Depressive symptoms. Wave 1 and 2 depressive symptoms were measured using a modified, 19-item version of the Centre of Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977). The Waves 3 and 4 Add Health interview data only included nine CES-D items. Shortened versions of the CES-D are frequently used by researchers (Roberts, Lewinsohn, & Seeley, 1991), as they require less time to complete and reduce respondent burden while retaining the psychometric properties of the original CES-D (Kohout, Berkman, Evans, & Cornoni-Huntley, 1993). Consistent with other studies using latent growth curve modeling, which require the use of functionally equivalent measures over time, the current study used nine CES-D items only (e.g., Meadows, Brown, & Elder, 2006). Respondents reported how often in the past seven days they have experienced each item: (1) you were bothered by things that usually don't bother you; (2) you felt that you were just as good as other people; (3) you could not shake off the blues, even with help from your family and your friends; (4) you had trouble keeping your mind on what you were doing; (5) you were depressed; (6) you were too tired to do things; (7) you enjoyed life; (8) you were sad; and (9) you felt that people disliked you. The range of responses varied from 0 (never or rarely) to 3 (most of the time). Items (2) and (7) were reverse coded so that higher scores reflected more severe depressive symptoms. Scores were summed to create a total score ranging from 0 to 27, with higher scores reflecting greater severity of depressive symptoms. The measure has demonstrated to be consistent, with Cronbach's alpha ranging from .77 to .82 across the four waves. Engagement in physical activity. Physical activity was measured by a modified version of a previously described scale (Ford, Nonnemaker, & Wirth, 2008), containing a range of self-report questions similar to those validated in other studies examining physical activity (Anderson, Crespo, Bartlett, Cheskin, & Pratt, 1998; Heath, Pratt, Warren, & Kann, 1994). In Wave 1 and 2, three items that assessed frequency of participation in the following activities during the past 7 days: (1) rollerblading, roller-skating, skateboarding, and bicycling; (2) playing an active sport, such as baseball, softball, basketball, soccer, swimming, or football; and (3) doing exercise, such as jogging, walking, karate, jumping rope, gymnastics, or dancing. Responses range from 0 (not at all) to 3 (5 or more times). Wave 3 and 4 included five slightly modified questions regarding participation in the following activities during the past 7 days: (1) bicycling, skateboarding, hiking, hunting, or doing yard work; (2) rollerblading, roller-skating, downhill skiing, snowboarding, playing racquet sports, or doing aerobics; (3) strenuous team sports such as playing football, soccer, basketball, lacrosse, rugby, field hockey, or ice hockey; (4) individual sports such as running, wrestling, swimming, cross country skiing, cycle racing, or martial arts; (5) participating in gymnastics, weight lifting, or strength training; (6) golfing, fishing or bowling, or playing softball or baseball; and (7) walking for exercise. Responses range from 0 (not at all) to 7 (7 or more times). In Waves 3 and 4, participants reported the frequency of the activities from Waves 1 and 2, plus additional activities that are developmentally relevant to young adults (e.g., weight lifting). Without including these additional developmentally relevant activities in Waves 3 and 4, frequency of physical activity during these waves may have appeared erroneously low, as emerging adults likely engage in different forms of physical activity compared to adolescents (e.g., strength training vs. skateboarding). Further, in order to avoid reporting inaccurately overestimated levels of physical activity in emerging adults as a result of the supplementary activities, a scaled sum of the frequency of engagement in physical activity was created (Gordon-Larsen, Nelson, & Popkin, 2004). The total sum of physical activity reported in Wave 3 and 4 was scaled to make it comparable to that of Waves 1 and 2. This was achieved by dividing by the sum of all activities

86

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

listed in the items presented at Wave 3 and 4 and then multiplying by the sum of all activities listed in the items presented at Wave 1 and 2 (Gordon-Larsen et al., 2004). Higher scores indicated greater frequency of engagement in physical activity. Gender. The gender variable was constructed from the Wave 1 in-home interview. Gender was dummy coded so that 0 ¼ female and 1 ¼ male. Covariates. Similar to previous studies (e.g., Birkeland et al., 2009; Jerstad et al., 2010; McPhie & Rawana, 2012; Raudsepp & Neissaar, 2012), body mass index (BMI) and socioeconomic (SES) status were included as covariates, as their influence was not of primary interest in the present study. BMI was calculated by dividing the participant's self-reported weight (in pounds) by squared height (in inches) and then multiplying by 703 (Centers for Disease Control and Prevention, 2003). SES was estimated using parent reported average household income. In the current sample, baseline SES was significantly, negatively associated with baseline depressive symptoms and significantly, positively associated with engagement in physical activity, whereas baseline BMI was not significantly associated with either baseline depressive symptoms or baseline physical activity levels. Data analysis In addition to descriptive statistics, latent growth curve modeling (LGCM) was used to achieve the chief study objectives. All analyses were carried out using AMOS version 19 and SPSS version 19. First, a normative growth trajectory was constructed by using data at the four time-points to estimate an unconditional growth model (UGM) for depressive symptoms. The factor loadings for the four measures of depressive symptoms were fixed to 1 on the intercept factor to signify that the intercept is invariant across time. Since it was expected that there would be a non-linear growth curve of depressive symptoms, the factor loadings relating the four measures of depressive symptoms to the slope factor were freely estimated in order to more flexibly capture the shape of the growth curve over the four time points (Bollen & Curran, 2006). In this case, only the first (fixed to 0) and the last (fixed to 1) factor loadings were constrained. In the final model, the freely estimated factor loadings for the second and third factors were 0.108 and 1.243, respectively. After constructing the UGM, a subsequent multivariate latent conditional model involving explanatory variables was estimated by adding in the time-varying predictor (i.e., physical activity). The growth in the time-varying predictor itself was modeled before assessing the effect of the time-varying predictor on growth in depressive symptoms (Bollen & Curran, 2006). To address the first study objective, the latent intercept and latent slope factor were estimated for depressive symptoms as well as engagement in physical activity using a freed-factor loading approach. Next, in order to model the observed variability in the intercept and slope of the depressive symptoms trajectory, the depressive symptom latent growth factor was regressed on the intercept and slope of engagement in physical activity. Lastly, in order to address the second and third goals of the study, the main effect and interaction of gender, a multigroup structural equation modeling (SEM) approach was used. Model fit was assessed using three fit indices: root-mean-square error of approximation (RMSEA), comparative fit index (CFI), and TuckereLewis index (TLI; Hu & Bentler, 1999; MacCallum & Austin, 2000). Missing data in the present study was handled using the full information strategy, which involves estimating model parameters using all available information, irrespective of whether that information comes from cases with incomplete data (Preacher, 2010). Results Descriptive statistics are provided in Table 1. A correlation matrix for the main study variables is presented in Table 2. Correlations across the four time-points were (r's ¼ .30e.56) for depressive symptoms and (r's ¼ .46e.67) for physical activity. Bivariate correlations between physical activity and depressive symptoms were in the negative direction. Table 1 Sample characteristics at baseline (N ¼ 3676). Full sample

Girls

Boys

M (SD) or N (%) for categorical variables

M (SD) or N (%) for categorical variables

45,492 (44,427)

1893 (51%) 45,493 (49,234)

1783 (49%) 45,490 (38,845)

1913 (52.0%) 781 (21.2%) 109 (3.0%) 209 (5.7%) 616 (16.8%) 46 (1.3%) 22.54 (4.44)

13.14e56.38

954 (50.4%) 437 (23.1%) 53 (2.8%) 104 (5.5%) 317 (16.7%) 27 (1.4%) 22.51 (4.48)

959 (53.8%) 344 (19.3%) 56 (3.1%) 105 (5.9%) 299 (16.8%) 19 (1.1%) 22.32 (4.40)

6.04 (4.29)

0.0e27.0

6.86 (4.66)

5.16 (3.65)

7.85 (5.03)

0.0e21.0

6.68 (4.61)

9.10 (5.15)

M (SD) or N (%) for categorical variables Demographic Variables Gender Household income (in dollars) Race/Ethnicity White Black Native American Asian/Pacific Islander Hispanic Other BMI Outcome Variable Depressive symptoms Predictor Variable Physical activity

Range

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

87

Unconditional growth model (UGM) The free-loading UGM for depressive symptoms was estimated and was found to fit the data well, with related fit statistics of CFI ¼ .98, TLI ¼ .94, and RMSEA ¼ .057 (.042e.073). Overall, respondents experienced a decrease in depressive symptoms across the transition from adolescence to EA (mean intercept ¼ 6.06, p < .001; mean slope ¼ 1.01, p < .001), suggesting relative intraindividual variability in depressive symptoms over development. There was a significant (p < .001) estimate for the covariance between the intercept and slope factors (covariance of slope and intercept ¼ 3.85) that was standardized to a correlation of r ¼ .60, indicating that adolescents reporting higher initial levels of depressive symptomatology tended to report a steeper decrease in depressive symptomatology over time. The depressive symptoms trajectory is depicted by notable variation in both the intercept (variance of the intercept ¼ 10.25, p < .001) and slope factors (variance of the slope ¼ 4.04, p < .001), suggesting substantial individual variability around both the mean baseline and mean rate of change over time. Given the presence of variability in interindividaul differences in the developmental growth trajectory of depressive symptoms, the next logical step is to ascertain whether engagement in physical activity corresponds to a source of this variation. The free-loading UGM for engagement in physical activity was estimated and found to fit the data well, with related fit statistics of CFI ¼ .99 TLI ¼ .95 and RMSEA ¼ .037 (.022e.055). In general, respondents experienced a decrease in engagement in physical activity across the transition from adolescence to EA (mean intercept ¼ 7.86, p < .001; mean slope ¼ 4.69, p < .001). There was a significant (p < .001) estimate for the covariance between the intercept and slope factors (covariance of slope and intercept ¼ 8.09) that was standardized to a correlation of r ¼ .84, reflecting that adolescents experiencing higher initial levels of engagement in physical activity tended to experience steeper declines in engagement in physical activity over time. The engagement in physical activity trajectory is depicted by notable variation in both the baseline (variance of the intercept ¼ 11.23, p < .001) and the rate of change (variance of the slope ¼ 8.09, p < .001). Growth model with time-varying predictor In order to model the observed variability in the intercept and slope factors of the depressive symptom trajectory, the time-varying predictor (i.e., physical activity) was added to the model, while adjusting for BMI and SES. The engagement in physical activity slope did not significantly predict the slope of depression symptoms. As such, a more restricted model was generated by omitting this pathway. This simplified model was estimated and found to fit the data just as well as the more complex model (i.e., model with more free parameters) as indicated by the non-significant chi-square difference test, Dc2 (1) ¼ .285, p > .05. In this case, the more parsimonious model (i.e., model with fewer free parameters) was retained for further analyses. The results presented in Table 3 (model 1) are from a multivariate growth model in which the factors of depressive symptoms are regressed on the intercept and slope of engagement in physical activity. The model was found to fit the data well. Engagement in physical activity was found to significantly predict the initial status of depressive symptoms (b ¼ 0.21, p < .001) and the rate of change in depressive symptoms (b ¼ 0.11, p < .001). Thus, those individuals with higher engagement in physical activity had, on average, lower levels of depressive symptoms at baseline. Further, high levels of engagement in physical activity at baseline led to a more gradual decline in depressive symptomatology across development (see Fig. 1). This time-varying predictor accounted for 6.1% of the variance in the depressive symptom intercept and 3.1% of the variance in the depressive symptom slope. Growth model with time-varying predictor and gender To address the hypotheses related to gender, a multigroup structural equation modeling approach was used to compare boys and girls on structural parameters related to physical activity and depressive symptoms. The results are presented in Table 3 (model 2). First, comparisons were made between the two unconstrained multivariate growth models for boys and girls. This model, in which the parameters were allowed to be unequal across genders, fit the data well (CFI ¼ .98; TLI ¼ .95; RMSEA ¼ .019, .015 to .023). No significant differences were found between the measurement weights, c2 (4) ¼ 4.54, p ¼ .34, or the structural weights, c2 (14) ¼ 17.83, p ¼ .22, for the boy and girl models, suggesting that they are equivalent across Table 2 Correlation matrix of the relationship between depressive symptoms and physical activity across waves (W). 1 1. 2. 3. 4. 5. 6. 7. 8.

Depressive symptoms Depressive symptoms Depressive symptoms Depressive symptoms Physical activity W1 Physical activity W2 Physical activity W3 Physical activity W4

W1 W2 W3 W4

*p < .05.**p < .01.***p < .001.

2

3

4

5

6

7

e .06** .04* .02 .06**

e .42*** .22*** .17***

e .25*** .20***

e .34***

e .56*** .33*** .30*** .12*** .12*** .05* .07***

e .36*** .34*** .08*** .09*** .03 .04**

e .38*** .03 .04 .02 .03

88

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

Table 3 Predictors of initial level and rate of change in depressive symptomatology using multivariate growth models. Predictors and summary statistics

Depressive symptomatology Model 1

Model 2

Intercept

Slope

0.21 (0.03)*** .061 .022 (.016e.028) .99 .97

Physical activity R2 RMSEA CFI TLI

Girls

0.11 (0.03)*** .031

Boys

Intercept

Slope

Intercept

Slope

0.15 (0.03)*** .029 .017 (.014e.021) .97 .96

0.07 (0.03)** .01

0.15 (0.03)*** .049 .017 (.014e.021) .97 .96

0.07 (0.03)** .014

**p  .01. ***p < .001. Note. Unstandardized coefficients presented with standard errors are in parentheses. Model 1 is the multivariate LGCM with the time-varying predictor. Model 2 is the multigroup multivariate LGCM with the time-varying predictor.

10 9 8

Depression Score

7 6 5 4 3

High engagement in physical activity

2

Low engagement in physical activity

1 0 13

15

17

19

21 Age in Years

23

25

27

29

Fig. 1. Model predicted growth trajectories for depressive symptoms as a function of frequency of engagement in physical activity during mid-adolescence.

groups. Between the two groups, significant differences were found for the structural intercepts, c2 (18) ¼ 458.74, p < .000, and the structural means, c2 (20) ¼ 460.89, p < .000. Follow-up analyses were conducted in order to determine which means and intercepts significantly differed across groups. First, the measurement weights and structural weights were fixed to be equal for both groups. This model in which the structural and measurement weights were constrained to be equal across groups fit the data well (CFI ¼ .97; TLI ¼ .96; RMSEA ¼ .017, .014 to .021). Next, one-by-one, each structural intercept or structural mean was fixed to be equivalent across groups. The chi-square statistic for each of these follow-up comparisons was contrasted to the chi-square statistic of the original model in order to see whether significant differences existed. These follow-up analyses showed a significant difference for all four latent variable means. Specifically, significant differences from the original model, c2 (70) ¼ 148.19, p ¼ .000, were found for boys and girls in terms of the depression intercept, Dc2 (1) ¼ 77.2, p < .001, physical activity intercept, Dc2 (1) ¼ 258.57, p < .001, depression slope, Dc2 (1) ¼ 24.73, p < .001, and physical activity slope, Dc2 (1) ¼ 64.64, p < .001. Thus, the models that fixed the conditional means to be equal across groups fit significantly worse than the original model, which allowed these parameters to be unequal across groups, suggesting gender differences (see Fig. 2). Adolescent boys (mean intercept ¼ 6.88, p < .001; mean slope ¼ 1.14, p < .01) had lower initial levels of depressive symptoms and slower rates of change in depressive symptoms compared to girls (mean intercept ¼ 8.13, p < .001; mean slope ¼ 1.84, p < .001). Adolescent boys (mean intercept ¼ 8.19, p < .001; mean slope ¼ 4.87, p < .001) had higher initial levels of physical activity and faster rates of change in frequency of physical activity compared to girls (mean intercept ¼ 5.84, p < .001; mean slope ¼ 3.62, p < .001). As previously reported, the structural weights were not found to differ between girls and boys, suggesting no interaction. In other words, the regression coefficients were found to be the same for girls and boys. As such, net of gender, BMI, and SES, engagement in physical activity remained a significant predictor of initial levels of depressive symptomatology (b ¼ 0.15, p < .001). Initial levels of engagement in physical activity remained a significant predictor of the rate of change in depressive

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

89

9 8

Depression Score

7 6 5 4 3 2

Boys Girls

1 0 13

15

17

19

21

23

25

27

29

Age in Years Fig. 2. Model predicted growth trajectories for depressive symptoms as a function of gender.

symptoms (b ¼ 0.07, p < .01). The set of predictors accounts for 4.9% of the variance in the depressive symptoms intercept for boys, and 2.9% for girls, and 1.4%% of the variance in the depressive symptom slope for boys, and 1.0% for girls. This model will be used as the final model for interpretation. Discussion Depressive symptom and physical activity trajectories The objectives of this study was to examine the longitudinal trajectory of depressive symptoms and engagement in physical activity in a national sample of adolescents and emerging adults, and to explore the predictive effect of engagement in physical activity and gender on this trajectory. Overall, the findings indicate that depressive symptoms generally decrease from adolescence into EA. This finding is notable given the limited research on changes in depressive symptoms spanning adolescence through to late EA (Wight, Sepu;lveda, & Aneshensel, 2004), as well as the limited research regarding depression trajectories in single cohorts of adolescents and emerging adults (Galambos, Barker, & Krahn, 2006). This result is consistent with the majority of past literature examining longitudinal trends in depressive symptomatology (e.g., Adkins Wang, Dupre, van den Oord, & Elder, 2009; Galambos et al., 2006; Marmorstein, 2009; Needham, 2007). Additionally, this trend is supported by past literature demonstrating that adolescents face the most elevated risk for depression onset around 15e18 years of age (Hankin et al., 1998). With regard to the trajectory of physical activity, rates of engagement in physical activity were found to decline during the transition from adolescence into EA. This finding is supported by past longitudinal studies (Birkeland et al., 2009; Gordon-Larsen et al., 2004; Kwan, Cairney, Faulkner, & Pullenayegum, 2012), which have found a similar trend of declining rates of physical activity during the transition into young adulthood. Predictors of the depressive symptom trajectory Overall, the results suggest that the severity of depressive symptoms among mid-adolescents is influenced by rates of engagement in physical activity. The salience of engaging in physical activity as a protective factor suggests that adolescents who engage in higher frequencies of physical activity are more resilient to developing depressive symptoms (see Fig. 1). This finding is consistent with past research that has identified an inverse relationship between physical activity and depression (Elliot et al., 2012; Jerstad et al., 2010; McPhie & Rawana, 2012). It was hypothesized that level of physical activity during midadolescence would be predictive of steeper decreases in depression symptoms and more gradual inclines in depressive symptoms during the transition from adolescence into EA. The predictive effect of physical activity on the depressive symptom trajectory was confirmed; although, contrary to what was expected, high levels of physical activity during midadolescence predicted more gradual declines in depressive symptoms over time. Although, adolescents with high levels of physical activity during mid-adolescence demonstrated slower declines in depressive symptoms, their depressive symptomatology remained consistently lower than their peers who engaged in lower levels of physical activity at age 15. Further, adolescents with high engagement in physical activity at age 15 might not have demonstrated the expected rapid declines in depressive symptoms over time as a result́of their already low levels of depressive symptomatology.

90

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

Although no known study has examined the long-term relationship between physical activity and depressive symptomatology from adolescence into late EA, longitudinal research using adult samples have reported mixed findings, while studies using adolescents have found some support for a prospective relationship. The findings in the current study are consistent with other studies (e.g., Jerstad et al., 2010; Sagatun, Søgaard, Bjertness, Selmer, & Heyerdahl, 2007) that have found a protective effect of physical activity on later depressive symptoms but contradict the findings of Birkeland et al. (2009). The present longitudinal study, which examined this relationship over 14 years, captured prospective systematic long-term relationships between level of engagement in physical activity during mid-adolescence and rate of change in depressive symptoms in a large, national sample of adolescent boys and girls. This suggests that engagement in physical activity serves as a protective factor, buffering against the experience of depression, possibly by fostering social connectedness, increasing self-esteem and efficacy, providing an adaptive outlet for managing one's emotions, or by directly operating on selected mechanisms and structures within the brain (Hillman & Drobes, 2012; Jerstad et al., 2010; McPhie & Rawana, 2012). Gender differences were also explored in terms of the depressive symptom trajectory. It was expected that adolescent girls would have higher levels of depressive symptoms at mid-adolescence compared to adolescent boys (see Fig. 2). This hypothesis was confirmed and is supported by the existing literature on gender differences in depression (Galambos et al., 2006; Garber, Keiley, & Martin, 2002; Hyde et al., 2008; Twenge & Nolen-Hoeksema, 2002). As expected, significant gender differences were found regarding the rate of change in depressive symptoms across development. Although boys begin and end the developmental period assessed in the study with less severe depressive symptoms than girls, boys were found to experience more gradual decreases in depressive symptoms overtime. This suggests that as adolescents undergo the transition into EA, the gender difference in depressive symptomatology begins to narrow (Needham, 2007). This reversal of the gender gap that typifies trajectories of depressive symptoms in adolescence has also been observed in other studies (e.g., Galambos et al., 2006; Needham, 2007). Further, it has been reported that the gender gap is narrowest between ages 18 and 30 and then begins to increase again (Mirowsky, 1996). Thus, the results suggest that EA represents a period in which the mental well-being of young people is at a relatively optimal level and that the gender gap that characterizes trajectories of depression during adolescence could once again reappear as women and men are faced with various transitions and life stressors that might influence them differentially (Galambos et al., 2006). The present study found no evidence for a gender-by-physical activity interaction in relation to the depressive symptom trajectory. It may be that engagement in physical activity has similar benefits for both genders with respect to its influence on severity of depressive symptoms across development. Future studies should aim to replicate this finding. Study limitations and future considerations The current research possess several strengths, including the use of a large, longitudinal sample of girls and boys; a prospective design; the application of LGCM to determine the trajectory of depressive symptomatology; the inclusion of the recently released Wave 4 of the Add Health data, which allowed for the examination of non-linear trajectories; and the extension of the depressive symptom trajectory into late EA. Despite these strengths, some study limitations exist. First, the physical activity measure only queried about frequency and not duration of physical activity or total energy expenditure, which could have an influence on depressive symptoms. Related, the different socioenvironmental contexts of physical activity are not considered. For example, some youth may be engaging in competitive sport programs (e.g., high school or university/college teams), while others may be engaging in non-competitive activities (e.g., hiking, biking, etc.) or structured exercise activities (e.g., fitness classes). Second, study findings might not generalize to other populations, such as non-schoolbased subpopulations and individuals experiencing clinical levels of depression. Third, it may be limiting to presume a single common trajectory for depressive symptoms during adolescence through EA (Raudenbush, 2001), especially given recent research suggesting that multiple distinct development pathways might exist (Brendgen, Wanner, Morin, & Vitaro, 2005; Rodriguez, Moss, & Audrain-McGovern, 2005). Forthcoming research should strive to address some of these limitations. Finally, the bivariate correlations among depressive symptomatology and physical activity were relatively weak. This may be due to a number of reasons. For one, depressive symptoms may be more strongly associated with other aspects of physical activity, such as duration or intensity, rather than frequency. Second, the range of values for depressive symptoms and physical activity may be restricted given the use of a normative sample. Lastly, weak correlations may result from a non-linear relation between physical activity and depressive symptoms. Study implications This study has made an original contribution to the literature by investigating the protective role of physical activity on the trajectory of depressive symptoms extending from adolescence into late EA. The results are relevant to developing programs and services intended to promote adaptive psychological functioning before the onset of elevated risk for depression. The results highlight the importance of engagement in physical activity in promoting mental well-being among adolescents and emerging adults. Moreover, the findings suggest that engagement in physical activity is not only important for reducing obesity among youth (e.g., Janssen, Katzmarzyk, Boyce, King, & Pickett, 2004), but also for mental health issues, such as depression.

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

91

References Adkins, D. E., Wang, V., Dupre, M. E., van den Oord, E. J. C. G., & Elder, G. H., Jr. (2009). Structure and stress: trajectories of depressive symptoms across adolescence and young adulthood. Social Forces, 88, 31e60. http://dx.doi.org/10.1353/sof.0.0238. Anderson, R. E., Crespo, C. J., Bartlett, S. J., Cheskin, L. J., & Pratt, M. (1998). Television watching with body weight and level of fatness among children: results from the Third National Health and Nutrition Examination Survey. Journal of American Medical Association, 279, 938e942. http://dx.doi.org/10.1001/ jama.279.12.938. Arnett, J. J. (2000). Emerging adulthood: a theory of development from the late teens through twenties. American Psychologist, 55, 469e480. http://dx.doi. org/10.1037/0003-066X.55.5.469. Birkeland, M. S., Torsheim, T., & Wold, B. (2009). A longitudinal study of the relationship between leisure-time physical activity and depressed mood among adolescents. Psychology of Sport and Exercise, 10, 25e34. http://dx.doi.org/10.1016/j.psychsport.2008.01.005. Bollen, K. A., & Curran, P. J. (2006). Latent curve models: A structural equation perspective. Hoboken, New Jersey: John Wiley & Sons. Brand, S., Gerber, M., Beck, J., Hatzinger, M., Pühse, U., & Holsboer-Trachsler, E. (2010). High exercise levels are related to favorable sleep patterns and psychological functioning in adolescents: a comparison of athletes and controls. Journal of Adolescent Health, 46, 133e141. http://dx.doi.org/10.1016/j. jadohealth.2009.06.018. Brendgen, M., Wanner, B., Morin, A., & Vitaro, F. (2005). Relations with parents and peers, temperament, and trajectories of depressed mood during early adolescence. Journal of Abnormal Child Psychology, 33, 579e594. http://dx.doi.org/10.1007/s10802-005-6739-2. Centers for Disease Control and Prevention. (2003). Retrieved July 4, 2011, from http://www.cdc.gov/healthyweight/assessing/bmi/adult_bmi/index.html. Cicchetti, D., & Toth, S. L. (1998). The development of depression in children and adolescents. American Psychologist, 53, 221e241. http://dx.doi.org/10.1037/ 0003-066X.53.2.221. Elliot, C. A., Kennedy, C., Morgan, G., Anderson, S. K., & Morris, D. (2012). Undergraduate physical activity and depressive symptoms. American Journal of Behaviors, 36, 230e241. http://dx.doi.org/10.1111/j.1476-5381.2012.01892.x. Ford, C. A., Nonnemaker, J. M., & Wirth, K. E. (2008). The influence of adolescent body mass index, physical activity, and tobacco use on blood pressure and cholesterol in young adulthood. Journal of Adolescent Health, 43, 576e583. http://dx.doi.org/10.1016/j.jadohealth.2008.06.010. Galambos, N. L., Barker, E. T., & Krahn, H. J. (2006). Depression, self-esteem, and anger in emerging adulthood: seven-year trajectories. Developmental Psychology, 42, 350e365. http://dx.doi.org/10.1037/0012-1649.42.2.350. Garber, J., Keiley, M. K., & Martin, N. C. (2002). Developmental trajectories of adolescents' depressive symptoms: predictors of change. Journal of Consulting and Clinical Psychology, 70, 79e95. http://dx.doi.org/10.1037//0022-006X.70.1.79. Gordon-Larsen, P., Nelson, M. C., & Popkin, B. M. (2004). Longitudinal physical activity and sedentary behavior trends: adolescence to adulthood. American Journal of Preventative Medicine, 27, 277e283. http://dx.doi.org/10.1016/j.amepre.2004.07.006. Hankin, B. L., Abramson, L. Y., Mofitt, T. E., Silva, P. A., McGee, R., & Angell, K. E. (1998). Development of depression from preadolescence to young adulthood: emerging gender differences in a 10-year longitudinal study. Journal of Abnormal Psychology, 107, 128e140. http://dx.doi.org/10.1037//0021-843X.107.1. 128. Harris, K. Mullan (2009). The National Longitudinal Study of Adolescent Health (Add Health), waves I & II, 1994e1996; wave III, 2001e2002; wave IV, 2007e2009 [machine-readable data file and documentation]. Chapel Hill, NC: Carolina Population Center, University of North Carolina at Chapel Hill. Heath, G. W., Pratt, M., Warren, C. W., & Kann, L. (1994). Physical activiry patterns in American high school student. Results from the 1990 Youth Risk Behavior Survey. Archives of Pediatric & Adolescent Medicine, 148, 1121e1136. http://dx.doi.org/10.1001/archpedi.1994.02170110017003. Hillman, C. H., & Drobes, D. J. (2012). Physical activity and cognitive control: Implications for drug abuse. Child Development Perspectives, 6, 367e373. http:// dx.doi.org/10.1111/cdep.12000. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1e55. http://dx.doi.org/10.1080/10705519909540118. Hyde, J. S., Mezulis, A. H., & Abramson, L. Y. (2008). The ABCs of depression: Integrating affective, biological, and cognitive models to explain the emergence of gender differences in depression. Psychological Review, 115, 291e313. http://dx.doi.org/10.1037/0033-295X.115.2.291. Janssen, I., Katzmarzyk, P. T., Boyce, W. F., King, M. K., & Pickett, W. (2004). Overweight and obesity in Canadian adolescents and their associations with dietary habits and physical activity patterns. Journal of Adolescent Health, 35, 360e367. http://dx.doi.org/10.1016/j.jadolhealth.2003.11.095. Jerstad, S. J., Boutelle, K. N., Ness, K. K., & Stice, E. (2010). Prospective reciprocal relations between physical activity and depression in female adolescents. Journal of Consulting and Clinical Psychology, 78, 268e272. http://dx.doi.org/10.1037/a0018793. Joormann, J., & D'Avanzato, C. (2010). Emotion regulation in depression” Examining the role of cognitive processes. Cognition and Emotion, 24, 913e939. http://dx.doi.org/10.1080/02699931003784939. Josefsson, T., Lindwall, M., & Archer, T. (2014). Physical exercise intervention in depressive disorders: meta-analysis and systematic review. Scandinavian Journal of Medicine & Science in Sports, 24, 259e272. http://dx.doi.org/10.1111/sms.12050. Kalak, N., Gerber, M., Kirov, R., Mikoteit, T., Yordanova, J., Pühse, U., et al. (2012). Daily morning running for 3 weeks improved sleep and psychological functioning in healthy adolescents compared with controls. Journal of Adolescent Health, 51, 615e622. http://dx.doi.org/10.1016/j.jadohealth.2012.02. 020. Kessler, R. C., Avenevoli, S., & Merikangas, K. R. (2001). Mood disorders in children and adolescents: an epidemiological perspective. Biological Psychiatry, 49, 1002e1014. http://dx.doi.org/10.1016/S0006-3223(01)01129-5. Kohout, F. J., Berkman, L. F., Evans, D. A., & Cornoni-Huntley, J. (1993). Two shorter forms of the CES-D Depression Symptom Index. Journal of Aging & Health, 5, 179e193. http://dx.doi.org/10.1177/089826439300500202. Kwan, M. Y., Cairney, J., Faulkner, G. E., & Pullenayegum, E. E. (2012). Physical activity and other health-risk behaviors during the transition into early adulthood: a longitudinal cohort study. American Journal of Preventative Medicine, 42, 14e20. http://dx.doi.org/10.1016/j.amepre.2011.08.026. €rjesson, M., & Ahlborg, G. (2014). The relationship of change in physical activity with change in depression, Lindwall, M., Gerber, M., Jonsdottir, I. H., Bo anxiety and burnout: a longitudinal study of Swedish healthcare workers. Health Psychology, 33, 1309e1318. http://dx.doi.org/10.1037/a0034402. MacCallum, R. C., & Austin, J. T. (2000). Applications of structural equation modeling in psychological research. Annual Review of Psychology, 51, 201e226. http://dx.doi.org/10.1146/annurev.psych.51.1.201. Marmorstein, N. R. (2009). Longitudinal associations between alcohol problems and depressive symptoms: early adolescence through early adulthood. Alcoholism: Clinical and Experimental Research, 33, 49e59. http://dx.doi.org/10.1111/j.1530-0277.2008.00810.x. McPhie, M. L., & Rawana, J. S. (2012). Unraveling the relation between physical activity, self- esteem and depressive symptoms among early and late adolescents: a mediation analysis. Mental Health & Physical Activity, 5, 43e49. http://dx.doi.org/10.1016/j.bbr.2011.03.031. Meadows, S. O., Brown, J. S., & Elder, G. H., Jr. (2006). Depressive symptoms, stress, and support: gendered trajectories from adolescence to young adulthood. Journal of Youth and Adolescence, 35, 93e103. http://dx.doi.org/10.1007/s10964-005-9021-6. Mirowsky, J. (1996). Age and the gender gap in depression. Journal of Health and Social Behavior, 37, 362e380. http://dx.doi.org/10.2307/2137263. Mota-Pereira, J., Silverio, J., Carvalho, S., Ribeiro, J. C., Fonte, D., & Ramos, J. (2011). Moderate exercise improves depression parameters in treatment-resistant patients with major depressive disorder. Journal of Psychiatry Research, 45, 1005e1011. http://dx.doi.org/10.1016/j.jpsychires.2011.02.005. Motl, R. W., Birnbaum, A. S., Kubik, M. Y., & Dishman, R. K. (2004). Naturally occurring changes in physical activity are inversely related to depressive symptoms during early adolescence. Psychosomatic Medicine, 66, 336e342. http://dx.doi.org/10.1097/?01.psy.0000126205.35683.0a. Needham, B. L. (2007). Gender differences in trajectories of depressive symptomatology and substance use during the transition from adolescence to young adulthood. Social Science & Medicine, 65, 1166e1179. http://dx.doi.org/10.1016/j.socscimed.2007.04.037. Needham, B. (2008). Reciprocal relationships between symptoms of depression and parental support during the transition from adolescence to young adulthood. Journal of Youth and Adolescence, 37, 893e905. http://dx.doi.org/10.1007/s10964-007-9181-7.

92

M.L. McPhie, J.S. Rawana / Journal of Adolescence 40 (2015) 83e92

Nolen-Hoeksema, S., & Girgus, J. S. (1994). The emergence of gender differences in depression during adolescence. Psychological Bulletin, 115, 424e443. http://dx.doi.org/10.1037/0033-2909.115.3.424. Paus, T., Keshavan, M., & Giedd, J. N. (2008). Why do many psychiatric disorders emerge during adolescence? Nature Reviews Neuroscience, 9, 947e957. http://dx.doi.org/10.1038/nrn2513. Preacher, K. J. (2010). The reviewer's guide to quantitative methods in the social sciences. In G. R. Hancock, & R. O. Mueller (Eds.), The reviewer's guide to quantitative methods in the social sciences (pp. 185e198). New York, NY: Routledge. Radloff, L. S. (1977). The CES-D scale: a self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385e401. http://dx.doi.org/10.1177/01466216770 R 0100306. Raudenbush, S. (2001). Comparing personal trajectories and drawing causal inferences from longitudinal data. Annual Reviews of Psychology, 52, 501e525. http://dx.doi.org/10.1146/annurev. psych.52.1.501. Raudsepp, L., & Neissaar, I. (2012). Brief report: relationship between physical activity and depressive symptoms in adolescent girls. Journal of Adolescence, 35, 1399e1402. http://dx.doi.org/10.1016/j.adolescence.2012.04.003. Roberts, R. E., Lewinsohn, P. M., & Seeley, J. R. (1991). Screening for adolescent depression: a comparison of depression scales. Child & Adolescent Psychiatry, 30, 58e66. http://dx.doi.org/10.1097/00004583-199101000-00009. Rodriguez, D., Moss, H., & Audrain-McGovern, J. (2005). Developmental heterogeneity in adolescent depressive symptoms: associations with smoking behavior. Psychosomatic Medicine, 67, 200e210. http://dx.doi.org/10.1097/01.psy.0000156929.83810.01. Rudolph, K. (2009). In I. Gotlib, & C. L. Hammen (Eds.), Handbook of depression (2nd ed.). (pp. 444e466). New York (NY): Guilford Press. Sagatun, A., Søgaard, A. J., Bjertness, E., Selmer, R., & Heyerdahl, S. (2007). The association between weekly hours of physical activity and mental health: a three-year follow-up study of 15-16-year-old students in the city of Oslo, Norway. BMC Public Health, 7, 115. http://dx.doi.org/10.1186/1471-2458-7-155. Sigfusdottir, I. D., Asgeirsdottir, B. B., Sigurdsson, J. F., & Gudjonsson, G. H. (2011). Physical activity buffers the effects of family conflict on depressed mood: a study on adolescent girls and boys. Journal of Adolescence, 34, 895e902. http://dx.doi.org/10.1016/j.adolescence. 2011. 01.003. Stanton, R., Happell, B., Hayman, M., & Reaburn, P. (2014). Exercise interventions for the treatment of affective disorders e research to practice. Frontiers in Psychiatry, 5, 46. http://dx.doi.org/10.3389/fpsyt.2014.00046. Twenge, J. M., & Nolen-Hoeksema, S. (2002). Age, gender, race, socioeconomic status, and birth cohort differences on the Children's Depression Inventory: a meta-analysis. Journal of Abnormal Psychology, 111, 578e588. http://dx.doi.org/10.1037//0021-843X.111.4.578. Vaughan, C. A., Foshee, V. A., & Ennett, S. T. (2010). Protective effects of maternal and peer support on depressive symptoms during adolescence. Journal of Abnormal Child Psychology, 38, 261e272. http://dx.doi.org/10.1007/s10802-009-9362-9. Wight, R. G., Sepulveda, J. E., & Aneshensel, C. S. (2004). Depressive symptoms: how do adolescents compare with adults? Journal of Adolescent Health, 34, 314e323. http://dx.doi.org/10.1016/j.jadohealth.2003.05.003. Yap, M. B. H., Allen, N. B., O'Shea, M., Di Parsia, P., Simons, J. G., & Sheeber, L. (2011). Early Adolescents' temperament, emotion regulation during motherchild interactions, and depressive symptomatology. Development and Psychopathology, 23, 267e282. http://dx.doi.org/10.1017/S095457941000078.

́