Adolescent mental health is an issue of global concern, with anxiety and depression symptoms being among the most widely reported complaints to impact wellbeing (Islam et al., Reference Islam, Khanam and Kabir2020; Merikangas et al., Reference Merikangas, He, Burstein, Swanson, Avenevoli, Cui, Benjet, Georgiades and Swendsen2010; Racine et al., Reference Racine, McArthur, Cooke, Eirich, Zhu and Madigan2021; Tiirikainen et al., Reference Tiirikainen, Haravuori, Ranta, Kaltiala-Heino and Marttunen2019). Critically, adolescence is a time of heightened stress and emotional reactivity and vulnerability for the emergence of psychopathology (Ahmed et al., Reference Ahmed, Bittencourt-Hewitt and Sebastian2015; Fuhrmann et al., Reference Fuhrmann, Knoll and Blakemore2015; Hauser et al., Reference Hauser, Will, Dubois and Dolan2019; Romeo, Reference Romeo2013; Shorey et al., Reference Shorey, Ng and Wong2021). It is a period of rapid development with extensive neural rewiring and psychosocial change (Andrews et al., Reference Andrews, Ahmed and Blakemore2021; Lichenstein et al., Reference Lichenstein, Verstynen and Forbes2016). In addition, normative developmental changes include shifts in adolescent sleep–wake behaviors that contribute to the prevalence of daytime sleepiness, which is a known risk factor for symptoms of anxiety and depression in adolescents (Crowley et al., Reference Crowley, Wolfson, Tarokh and Carskadon2018; Liu et al., Reference Liu, Zhang, Li, Chan, Yu, Lam, Chan, Li and Wing2019; Luo et al., Reference Luo, Zhang, Chen, Lu and Pan2018; Meyer et al., Reference Meyer, Barbosa, Junior, Andrade, Silva, Pelegrini and Gomes Felden2018).
Perceived Stress
There is a robust literature supporting links between stressors and child and adolescent psychopathology (Grant et al., Reference Grant, Compas, Thurm, McMahon and Gipson2004; Gunnar, Reference Gunnar2021). Early adolescence is marked by significant transitions in stress reactivity, as pubertal development and stress interact to impact one of the body’s main stress response systems — the hypothalamic–pituitary–adrenal (HPA) axis (Romeo, Reference Romeo2010). Indeed, studies in rats show that HPA axis function in response to stressors is prolonged in adolescence, compared to adulthood (McCormick & Mathews, Reference McCormick and Mathews2007). While most studies to date have focused on the effects of adverse life events (Grant et al., Reference Grant, Compas, Thurm, McMahon and Gipson2004; Gunnar, Reference Gunnar2021), perceived stress in relation to everyday stressors has also emerged as a marker for risk of developing a mental disorder (Lindholdt et al., Reference Lindholdt, Labriola, Andersen, Kjeldsen, Obel and Lund2021). School-related stressors have been identified as a major source of subjective stress in adolescents (Anniko et al., Reference Anniko, Boersma and Tillfors2019; Kaczmarek & Trambacz-Oleszak, Reference Kaczmarek and Trambacz-Oleszak2021). Perceived stress may offer insights that stress exposure measures do not, as perceived stress reflects an individual’s interpretation of environmental stressors and their anticipated ability to successfully manage stressful situations (Hu et al., Reference Hu, Koucky, Brown, Bruce and Sheline2014).
Rumination
Dysfunctional emotion regulation has been implicated as a key transdiagnostic construct in the development and maintenance of psychopathology (Olatunji et al., Reference Olatunji, Naragon-Gainey and Wolitzky-Taylor2013; Sloan et al., Reference Sloan, Hall, Moulding, Bryce, Mildred and Staiger2017; Watkins & Roberts, Reference Watkins and Roberts2020). In a metaanalytic review of emotion regulation strategies in adolescents aged 13 to 18 years, rumination was identified as a maladaptive strategy that negatively impacts subclinical adolescent anxiety and depression symptoms (Schafer et al., Reference Schafer, Naumann, Holmes, Tuschen-Caffier and Samson2017). Similar results have been reported for 722 younger adolescents, aged 11 to 13 years (Hilt et al., Reference Hilt, McLaughlin and Nolen-Hoeksema2010). Depressive rumination is a maladaptive thought process, or cognitive vulnerability, defined as repetitive self-focus on symptoms of depression, including possible causes and consequences, which can influence symptom duration and severity (Nolen-Hoeksema, Reference Nolen-Hoeksema1991; Watkins & Roberts, Reference Watkins and Roberts2020). Rumination has been associated with perceived stress and physiological stress responses (e.g. as evidenced by HPA axis indices) and further may mediate maladaptive HPA axis activation — a pathway through which it may influence mental health (Gianferante et al., Reference Gianferante, Thoma, Hanlin, Chen, Breines, Zoccola and Rohleder2014; Hilt, Sladek, et al., Reference Hilt, Sladek, Doane and Stroud2017; Hu et al., Reference Hu, Koucky, Brown, Bruce and Sheline2014). Adolescence is a particularly challenging time for the regulation of emotions as the underlying neural circuitry supporting these skills is still developing, while social demands and challenges are increasing (Ahmed et al., Reference Ahmed, Bittencourt-Hewitt and Sebastian2015; Andrews et al., Reference Andrews, Ahmed and Blakemore2021; Young et al., Reference Young, Sandman and Craske2019).
Daytime Sleepiness
Daytime sleepiness is an additional burden than can exacerbate adolescent mental health problems, and the relationship may be reciprocal (Hestetun et al., Reference Hestetun, Svendsen and Oellingrath2018; Luo et al., Reference Luo, Zhang, Chen, Lu and Pan2018). This may be due in part to influences on perceived stress and level of rumination, with follow-on effects for mental health (e.g. Bian et al., Reference Bian, Hou, Zuo, Quan, Ju, Wu and Xi2020; Lindholdt et al., Reference Lindholdt, Labriola, Andersen, Kjeldsen, Obel and Lund2021; Schafer et al., Reference Schafer, Naumann, Holmes, Tuschen-Caffier and Samson2017; Thomsen et al., Reference Thomsen, Yung Mehlsen, Christensen and Zachariae2003; Ye et al., Reference Ye, Wu, Wang, Im, Liu, Wang and Yang2021). For example, excessive daytime sleepiness has been associated with high levels of perceived study stress in high-school students (Luo et al., Reference Luo, Zhang, Chen, Lu and Pan2018). Sleepiness may worsen perceived stress levels by reducing an individual’s perceived ability to negotiate everyday stressors (Foster, Reference Foster2020). In addition, poor sleep quality has been associated with a tendency to ruminate in college students (Bian et al., Reference Bian, Hou, Zuo, Quan, Ju, Wu and Xi2020), and similarly, greater sleepiness has been associated with poorer emotional self-regulation in adolescents (Owens et al., Reference Owens, Dearth-Wesley, Lewin, Gioia and Whitaker2016). Importantly, daytime sleepiness is a very common adolescent complaint that has multiple etiologies, including pubertal maturation-induced changes in sleep architecture (Agostini & Centofanti, Reference Agostini and Centofanti2021; Hein et al., Reference Hein, Mungo, Hubain and Loas2020). As with stress and rumination, daytime sleepiness has been linked to HPA axis activation, and specifically, with blunted cortisol stress responsiveness (van Dalfsen & Markus, Reference van Dalfsen and Markus2018). This may reflect the chronic nature of daytime sleepiness, as blunted responsiveness is thought to characterize prolonged, or persistent stress exposure (Fries et al., Reference Fries, Hesse, Hellhammer and Hellhammer2005; Miller et al., Reference Miller, Chen and Zhou2007).
Longitudinal Associations
Longitudinal studies suggest that, in adolescents, mental health is prospectively influenced by both perceived psychological stress (Felton et al., Reference Felton, Banducci, Shadur, Stadnik, MacPherson and Lejuez2017; Lindholdt et al., Reference Lindholdt, Labriola, Andersen, Kjeldsen, Obel and Lund2021) and rumination (Orue et al., Reference Orue, Calvete and Padilla2014; Padilla Paredes & Calvete Zumalde, Reference Padilla Paredes and Calvete Zumalde2015; Wilkinson et al., Reference Wilkinson, Croudace and Goodyer2013). Stress has complex and reciprocal associations with anxiety and depression, and these relationships may be influenced by cognitive vulnerabilities, including rumination (Liu & Alloy, Reference Liu and Alloy2010; Michl et al., Reference Michl, McLaughlin, Shepherd and Nolen-Hoeksema2013; Morrison & O’Connor, Reference Morrison and O’Connor2005). Studies in adults suggest bidirectional associations between daytime sleepiness and depression (Jaussent et al., Reference Jaussent, Morin, Ivers and Dauvilliers2017; Karunanayake et al., Reference Karunanayake, Dosman, Fenton, Rennie, Kirychuk, Ramsden, Seeseequasis, Abonyi and Pahwa2019; LaGrotte et al., Reference LaGrotte, Fernandez-Mendoza, Calhoun, Liao, Bixler and Vgontzas2016). Similarly, sleep problems in twins aged 8 years predict depression at age 10, although the converse was not found (Gregory et al., Reference Gregory, Rijsdijk, Lau, Dahl and Eley2009).
Genetic Sources of Influence
Anxiety and depression, perceived stress, rumination, and daytime sleepiness are interrelated traits. All are associated with HPA axis function (Chen, Reference Chen2022; Gianferante et al., Reference Gianferante, Thoma, Hanlin, Chen, Breines, Zoccola and Rohleder2014; Romeo, Reference Romeo2010; van Dalfsen & Markus, Reference van Dalfsen and Markus2018), which may partly explain their covariation, and twin studies suggest genetic overlap among the traits. In young adults, covariation between perceived psychological stress and depressive symptoms appears to have a strong genetic component (Michalski et al., Reference Michalski, Demers, Baranger, Barch, Harms, Burgess and Bogdan2017; Rietschel et al., Reference Rietschel, Zhu, Kirschbaum, Strohmaier, Wust, Rietschel and Martin2014). Similar associations are reported between rumination and depressed mood in adolescents (Moore et al., Reference Moore, Salk, Van Hulle, Abramson, Hyde, Lemery-Chalfant and Goldsmith2013), while a modest overlap in genetic factors appears to contribute to covariation between daytime sleepiness and depressive symptoms in elderly men (Lessov-Schlaggar et al., Reference Lessov-Schlaggar, Bliwise, Krasnow, Swan and Reed2008).
The Current Study
This study aims to prospectively investigate the relative importance of perceived stress, rumination, and daytime sleepiness as predictors of adolescent anxiety and depression. Given the known associations between these traits, we set out to determine the extent to which they may independently predict anxiety/depression symptoms. Further, having a twin cohort provides the opportunity to explore the role of genetic and environmental influences on these relationships, providing insight into the nature of factors influencing persistence of anxiety/depression symptoms in early adolescence.
Methods
Participants
The Queensland Twin Adolescent Brain (QTAB) cohort was recruited primarily through the Queensland Twin Registry (QTwin), but with additional twins recruited through Twins Research Australia and through a QTAB study website (Strike et al., Reference Strike, Hansell, Chuang, Miller, de Zubicaray, Thompson, McMahon and Wright2022). All families, with young adolescent twins (or triplets), resided in south-east Queensland and were able to attend a brain imaging and data collection session at the Centre for Advanced Imaging, University of Queensland (for the 5 triplet sets, only 2 individuals per triplet set were imaged, and they are treated as twin pairs in all analyses). Baseline data were collected from 422 participants (211 families) between June 2017 and October 2019, with twin pairs aged 9.0 to 14.4 years (M = 11.3 ± 1.3 years). Twin pairs comprised 46 monozygotic female (MZF) pairs, 57 monozygotic male (MZM) pairs, 34 dizygotic female (DZF) pairs, 30 dizygotic male (DZM) pairs, and 44 opposite-sex (DZO) pairs. Of these, 152 families (including two families with triplets) participated in a second wave of data collection (November 2019 to January 2021). At follow-up, the twins were aged 10.1 to 16.4 years (M = 13.0 ± 1.5 years), with pairs comprising 36 MZF, 38 MZM, 26 DZF, 19 DZM, and 33 DZO pairs. The interval from baseline to follow-up ranged from 1.1 to 2.5 years (M = 1.7 ± .30 years).
Twin pairs were excluded if either co-twin had serious medical, neurological, or cardiovascular conditions, history of a serious head injury, diagnosis of autism spectrum disorder or psychiatric disorder, and any cognitive, physical, or sensory challenges that would limit ability to understand or complete procedures. In addition, MRI contraindication (e.g. metal implants, artifact-inducing orthodontic braces) was cause for exclusion. Written and informed consent was obtained from a parent and from each adolescent. A parent and each adolescent received an honorarium (AUD$50 each) for participating in the study. Ethics approval for the study was obtained from the Children’s Health Queensland Human Research Ethics Committee and ratified by The University of Queensland.
Procedure
Participants completed a mental health questionnaire on an iPad during their visit. Twins were assessed in parallel, with one twin being scanned while the other completed iPad and other assessments. Equivalent and additional assessments were completed by a parent. The nonimaging data are stored at The University of Queensland’s institutional repository, UQ eSpace (https://doi.org/10.48610/e891597). Access to the dataset can be requested via the Request access to the dataset link in the UQ eSpace record.
Measures
Predictor variables: perceived stress, rumination, daytime sleepiness
Perceived stress was obtained from the Daily Life Stressors Scale. This 30-item scale was designed to assess the impact of everyday problems and stressors, at school and at home, for children and adolescents aged 7−17 years, and this age group shows construct and concurrent validity (Kearney et al., Reference Kearney, Drabman and Beasley1993). The scale addresses the severity of negative daily life events (e.g. ‘My parents yell at me in the morning’) and negative affectivity (e.g. ‘I am tense or nervous when I have to answer a question in class’) within the past week. Items are self-rated on a 5-point Likert-type scale and summed to obtain a total score ranging from 0 to120. Where responses were missing for 1−2 items, we replaced with the participant’s mean item score (four participants were missing one item, and one participant was missing two items at baseline). Three participants missing nine or more responses at baseline were not scored. There were no missing data at follow-up.
Rumination was assessed with the Rumination Subscale of the Children’s Response Styles Questionnaire. This subscale comprises 13 items describing self-focused responses to depressed mood (Abela et al., Reference Abela, Brozina and Haigh2002) and has good validity and internal consistency in young adolescents (Alloy et al., Reference Alloy, Black, Young, Goldstein, Shapero, Stange, Boccia, Matt, Boland, Moore and Abramson2012; Shapero et al., Reference Shapero, McClung, Bangasser, Abramson and Alloy2017). Items are self-rated on a 4-point Likert scale and summed to obtain a total score ranging from 0 to 39. Baseline data were not available for three participants. There were no missing data at follow-up.
Daytime sleepiness was assessed with the Pediatric Daytime Sleepiness Scale, which has robust psychometric properties in adolescents aged 11 to 15 years (Drake et al., Reference Drake, Nickel, Burduvali, Roth, Jefferson and Pietro2003). Multiple studies have used the scale in younger children (see review by Meyer et al., Reference Meyer, Ferrari, Barbosa, Andrade, Pelegrini and Felden2017), including validation studies (e.g. Nouri et al., Reference Nouri, Esmaeili, Seyedi, Rezaeian, Panjeh, Cogo-Moreira and Pompeia2021). This self-report 8-item scale is scored on a 5-point Likert-type scale with items summed to obtain a total score ranging from 0 to 32. At baseline, missing data were replaced with the participant’s item mean for two participants missing one item. Three participants had no data at baseline for this scale. There were no missing data at follow-up.
Outcome variable: anxiety/depression
A measure of anxiety/depression symptoms was derived from the Somatic and Psychological Health Report (SPHERE). The SPHERE was first developed as a 34-item self-report questionnaire to assess symptoms of mental distress and persistent fatigue (Hickie et al., Reference Hickie, Davenport, Hadzi-Pavlovic, Koschera, Naismith, Scott and Wilhelm2001). Here we use a 14-item anxiety/depression subscale that has been validated in an Australian-based population sample aged 9 to 28 years, where it was associated with later DSM-IV diagnoses of major depressive disorder, social anxiety, and alcohol dependence (OR 1.23–1.47; Couvy-Duchesne et al., Reference Couvy-Duchesne, Davenport, Martin, Wright and Hickie2017). Similarly, it has been associated with concurrent neuroticism in adolescents and young adults (r = .64; Hansell et al., Reference Hansell, Wright, Medland, Davenport, Wray, Martin and Hickie2012). There were no missing data for this scale at baseline or follow-up. Items are self-rated on a 4-point Likert-type scale and summed to obtain a total score ranging from 0 to 42.
Covariates
Symptoms of anxiety and/or depression are higher in females than males (Altemus et al., Reference Altemus, Sarvaiya and Neill Epperson2014; Salk et al., Reference Salk, Hyde and Abramson2017) and increase with: age across adolescence (Salk et al., Reference Salk, Hyde and Abramson2017), advancing pubertal development (Altemus et al., Reference Altemus, Sarvaiya and Neill Epperson2014; Stumper & Alloy, Reference Stumper and Alloy2021), and socioeconomic disadvantage (Peverill et al., Reference Peverill, Dirks, Narvaja, Herts, Comer and McLaughlin2020). A potential confounder for prediction analyses is the varying interval between baseline and follow-up assessment. In addition, follow-up assessment was interrupted by a COVID-19-related lockdown of approximately 3 months, during which schools were closed. At follow-up, 31% of the sample was assessed prior to the lockdown, with the remainder assessed postlockdown.
Pubertal development was assessed using self-report and parental-report versions of the Pubertal Development Scale (Carskadon & Acebo, Reference Carskadon and Acebo1993; Petersen et al., Reference Petersen, Crockett, Richards and Boxer1988). This scale assesses the development of secondary sexual characteristics, with questions about growth spurts, body hair growth, and skin changes (e.g. pimples), as well as breast development and menarche in girls, and voice changes and facial hair growth in boys. Except for menarche, which is scored yes/no, items are scored on a 4-point Likert-type scale with options ranging from ‘has not begun’ to ‘already complete’. The average scale score (ranging from 1 to 4) was determined. At baseline, a pubertal scale score was available for all but one participant (supplemented with data from parent-report for 179 participants and scored on four rather than five items for one participant). At follow-up, a pubertal scale score was available for all but three participants (supplemented with data from parent-report for 83 participants and scored on four rather than five items for eight participants).
Neighborhood socioeconomic status (SES) was computed using the Australian Bureau of Statistics (ABS) 2016 Census-based Socioeconomic Index for Areas, which ranks suburbs in Australia according to relative socioeconomic advantage and disadvantage (ABS, 2016). Higher scores indicate greater advantage in general, including more high-income households and more people in skilled occupations. There were no missing data.
Analyses
Anxiety/depression, perceived stress, and rumination had minor positive skew and were square root-transformed then standardized (M = 0, SD = 1) and minor outliers (all <4SD) winsorized to ±3.3 SD. Analyses were conducted using IBM SPSS Statistics Version 27 and R version 4.1.2 (R Core Team, 2021). Unless stated otherwise, all analyses including baseline data use the full sample.
Phenotypic analyses
A series of linear regression models were used to explore cross-sectional and longitudinal associations.
Predicting anxiety/depression
Cross-sectional analyses examined, first, the extent to which perceived stress, rumination, and daytime sleepiness were associated with concurrent anxiety/depression. Second, they explored the independence of such associations with concurrent anxiety/depression.
Longitudinal analyses first determined that the proportion of variance in anxiety/depression at follow-up was predicted individually by baseline measures of daytime sleepiness, perceived stress, and rumination. The second series of longitudinal analyses included anxiety/depression at baseline as a predictor, to test whether perceived stress, rumination, or daytime sleepiness predicted change in anxiety/depression (i.e. residual variance in anxiety/depression at follow-up after accounting for baseline anxiety/depression). Third, relative to each other, the extent to which daytime sleepiness, perceived stress, and rumination predicted independent change in anxiety/depression was assessed.
Bidirectionality of influence across time
Using both baseline and follow-up measures, these analyses explored the extent of bidirectionality between anxiety/depression, perceived stress, rumination, and daytime sleepiness.
Mixed-effects linear regressions were conducted using the R package lme4 (Bates et al., Reference Bates, Mächler, Bolker and Walker2015). Covariates having at least a nominally significant association with the outcome variable were included as fixed effects. To account for the nonindependence of twins, family, and zygosity (identical vs. nonidentical) were included in the model as random effects (Visscher et al., Reference Visscher, Benyamin and White2004). An experiment-wide significance threshold of p < .0085, based on the identification of six effective independent predictor/outcome variables (equation 5 of Li & Ji, Reference Li and Ji2005), was adopted to keep the Type I error rate at 5%.
Genetic analyses
The genetic and environmental etiology of associations between baseline predictors (anxiety/depression, daytime sleepiness, perceived stress, and rumination) and follow-up anxiety/depression was examined using a twin study approach. In the classic twin design, structural equation modeling exploits twin relationships to determine quantitative parameter estimates of genetic and environmental contributions to trait variability (Rijsdijk & Sham, Reference Rijsdijk and Sham2002). The model assumes that monozygotic (MZ) twins share 100% of their genetic material, as they result from a single egg fertilized by a single sperm, while dizygotic (DZ) twins result from two eggs fertilized by two sperm, and share, on average, 50% of their genetic material. When twin pairs are raised together, it is possible to determine the impact of environmental influences that are shared by the twins (i.e. same family, school, neighborhood) as well as those that are unique to each twin, such that both MZ and DZ pairs share 100% of shared, or common, environmental influences, while unique environmental influences (including measurement error) are uncorrelated.
A series of bivariate Cholesky decompositions (Neale & Maes, Reference Neale and Maes1998) were used to decompose variance and covariance and, in particular, longitudinal associations between baseline predictors and follow-up anxiety/depression symptoms. Cholesky decomposition enabled the proportion of phenotypic association due to additive genetic (A), common/shared environmental (C), and unique or nonshared environmental (E) influences to be computed (Supplementary Figure S1). Prior to analysis, each variable was regressed on nominally significant covariates (i.e. p < .05), as identified in Supplementary Table S1 for the full baseline sample (i.e. age and neighborhood SES for anxiety/depression and perceived stress; sex and age for rumination) and Supplementary Table S2 for the sample at follow-up. Heritability for each variable at baseline and follow-up was derived from an 8-variable Cholesky decomposition. Analyses were conducted in R, with genetic analyses using OpenMx (Boker et al., Reference Boker, Neale, Maes, Wilde, Spiegel, Brick, Spies, Estabrook, Kenny, Bates, Mehta and Fox2011).
Results
Sample descriptive statistics are shown in Table 1. Note that while group means are similar, the highest scorers for anxiety/depression, perceived stress, and daytime sleepiness did not participate in the study at Time 2 (compare maximum scores for the full baseline sample and Time 1 longitudinal sample). Covariates are assessed in Supplementary Table S1 for variables Time 1 and Supplementary Table S2 for variables at Time 2. For the outcome variable (i.e. anxiety/depression at Time 2), only puberty and sex were nominally significant (Supplementary Table S2). Puberty accounted for approximately 5% of the variance in anxiety/depression at Time 2 (more advanced pubertal status = higher symptom level) and sex for approximately 2% of variance (higher symptom level in girls). Anxiety/depression symptoms were moderately stable from baseline to follow-up (r = 0.42, see Supplementary Table S3 for all correlations).
Table 1. Raw data descriptive statistics
Note: All analyses of baseline (Time 1) data use the full sample unless stated otherwise.
To What Extent Are Perceived Stress, Rumination, and Daytime Sleepiness Associated with Concurrent Measures of Anxiety/Depression? and Do They Have Any Independent Influence?
Perceived stress, rumination, and daytime sleepiness were moderately associated with concurrent symptoms of anxiety/depression at both baseline (Time 1) and follow-up (Time 2), where they accounted for 24% to 33% of variance (Supplementary Tables S4 and S5, Models 1−3), with 3% to 8% of this variance trait-specific.
Specifically, at Time 1, perceived stress is associated with 32% of the variance in anxiety/depression, of which 6% (comparing R 2 for Models 6 and 7, Supplementary Table S4) is independent of rumination and daytime sleepiness (33% and 3% respectively at Time 2, Supplementary Table S5). Rumination at Time 1 is associated with 24% of variance in anxiety/depression, of which 5% (comparing Models 5 and 7, Supplementary Table S4) is independent of perceived stress and daytime sleepiness (28% and 4% respectively at Time 2, Supplementary Table S5). Similarly, at Time 1, daytime sleepiness is associated with 24% of variance in anxiety/depression, of which 6% (comparing Models 4 and 7, Supplementary Table S4) is independent of perceived stress and rumination (33% and 8% respectively at Time 2, Supplementary Table S5).
Do Perceived Stress, Rumination, and Daytime Sleepiness Prospectively Predict Variance in Anxiety/Depression?
These results are shown in Tables 2–4 (Model 1). When assessed as the only predictor, perceived stress predicts approximately 11% of the variance in later anxiety/depression (i.e. R 2 = .106), while rumination and daytime sleepiness each predict approximately 8%.
Table 2. Results of linear regression models predicting anxiety/depression symptoms at Time 2 (T2) from perceived stress at Time 1 (T1)
Note: Family and zygosity are included as random effects. Model 5 shown in bold type is the best-fitting model. The inclusion of puberty T2 (Model 4) and sex (Model 5) do not result in beta weight drops for perceived stress T1 (i.e. compared to Model 3).
* p values less than the experiment-wide significance threshold (p < .0085).
Table 3. Results of linear regression models predicting anxiety/depression symptoms at Time 2 (T2) from rumination at Time 1 (T1)
Note: Family and zygosity are included as random effects. Model 5 shown in bold type is the best-fitting model. The inclusion of puberty T2 (Model 4) and sex (Model 5) results in beta weight drops for rumination T1, thus reflecting overlapping variance.
* p values less than the experiment-wide significance threshold (p < .0085).
Table 4. Results of linear regression models predicting anxiety/depression symptoms at Time 2 (T2) from daytime sleepiness at Time 1 (T1)
Note: Family and zygosity are included as random effects. Model 5 shown in bold type is the best-fitting model. The inclusion of puberty T2 (Model 4) results in a beta weight drop for daytime sleepiness T1, thus reflecting overlapping variance.
* p values less than the experiment-wide significance threshold (p < .0085).
Table 5. Results of linear regression models predicting anxiety/depression symptoms at Time 2 (T2) from perceived stress, rumination, and daytime sleepiness at Time 1 (T1)
Note: Family and zygosity are included as random effects in Models 1–8. Model 9 was overfitted if both family and zygosity were included and so was run with only family included as a random effect. Model 9 shown in bold type is the best-fitting model.
* p values less than the experiment-wide significance threshold (p < .0085).
Do Perceived Stress, Rumination, and Daytime Sleepiness Prospectively Predict Change in Anxiety/Depression?
Only perceived stress passes the significance threshold (i.e. p < .0085) when concurrent anxiety/depression is included in the model (Tables 2−4, Model 3). Approximately 3% of the variance in anxiety/depression at Time 2 is change predicted by perceived stress that is independent of anxiety/depression at Time 1 (comparing R 2 for Models 2 and 3, Table 2).
While the inclusion of puberty and sex accounts for additional variance and improves model fit (Models 4 and 5), their inclusion does not cause a drop in beta weight for perceived stress as identified in Model 3 (Table 2).
Do Perceived Stress, Rumination, and Daytime Sleepiness Independently Predict Later Anxiety/Depression?
Perceived stress, rumination, and daytime sleepiness together predict approximately 14% of variance in later anxiety/depression symptoms, with both perceived stress and daytime sleepiness accounting for significant independent variance (Table 5, Model 4). Perceived stress independently accounts for 2.6% of variance in later anxiety/depression (R 2 difference between Models 3 and 4, Table 5), while daytime sleepiness independently accounts for 2.1% (comparing Models 1 and 4). However, when concurrent anxiety/depression symptoms were added as a predictor (Table 5, Model 5), neither perceived stress nor daytime sleepiness remained significant. Therefore, while perceived stress as a single trait predicts change in later anxiety/depression symptoms (Table 2, Model 5), it does not do so independently of either daytime sleepiness or rumination at the experiment-wide significance threshold of p < .0085. The inclusion of puberty at Time 2 and sex (Table 5, Model 9) improves model fit and takes overall variance accounted for in anxiety/depression at Time 2 to approximately 24%.
Bidirectionality of Influences Across Time
As expected, anxiety/depression, daytime sleepiness, perceived stress, and rumination have significant bidirectional influences with each other (Figure 1, Supplementary Tables S6a–f). Perceived stress has a stronger influence on later anxiety/depression than vice versa (i.e. perceived stress at Time 1 accounted for 11% of variance in anxiety/depression at Time 2, while anxiety/depression at Time 1 accounted for 5% of variance in perceived stress at Time 2, Supplementary Table S6a, Models 1 and 4) and similarly for daytime sleepiness (8% vs. 3%, Supplementary Table S6c, Models 1 and 4). In contrast, levels of association between anxiety/depression and rumination over time were similar in both directions (8% vs. 7%, Supplementary Table S6b, Models 1 and 4). Bidirectional associations over time between perceived stress and rumination (7% vs. 12%, Supplementary Table S6d, Models 1 and 4), perceived stress and daytime sleepiness (9% vs. 6% Supplementary Table S6e, Models 1 and 4), and rumination and daytime sleepiness (7% vs. 7%, Supplementary Table S6f, Models 1 and 4) were of similar magnitude.
Fig. 1. Bidirectional influences over time, showing percentage of variance accounted for in the outcome variable at Time 2. For example, rumination at Time 1 accounts for 12% of variance in perceived stress at Time 2 (while perceived stress at Time 1 accounts for 7% of variance in rumination at Time 2). Each percentage represents the R 2 identified in a series of linear regressions (Supplementary Tables S6a–f, Models 1 and 4).
Genetic and Environmental Influence on Total Variation
While sample size is a limiting factor for our genetic analyses, we can nonetheless make some observations. In our adolescent sample, common environment and/or additive genetic sources account for approximately a third to half of the total variance in all traits (Figure 2, Supplementary Table S7). Remaining variance was influenced by unique environmental sources, which may include measurement error. Genetic sources of influence were significant for baseline anxiety/depression and perceived stress, rumination at follow-up, and for daytime sleepiness at both time points. Common environment had a significant influence on perceived stress and rumination at both time points, as well as anxiety/depression symptoms at follow-up. However, confidence intervals were broad for all significant additive genetic and common environmental influences, with the lower bound close to zero for many (Supplementary Table S7). Unique environmental influences were greater at baseline than follow-up for all traits, although confidence intervals overlap.
Fig. 2. Additive genetic (A), common environmental (C), and unique environmental influences on traits at Time 1 (T1) and Time 2 (T2) are shown as a percentage of total variance. Estimates are derived from multivariate analyses including all eight variables and using the full sample at Time 1. For results in the same sample at both timepoints, see Supplementary Figure S2 — this allows for better comparison, but using the larger Time 1 sample provides more accurate results overall. Nonsignificant estimates are notated ‘ns’. 95% confidence intervals are shown in Supplementary Table S7.
To What Extent Do Genetic and Environmental Factors Account for Associations Between Time 1 Predictors (Anxiety/Depression, Perceived Stress, Rumination, Daytime Sleepiness) and Symptoms of Anxiety/Depression at Time 2?
Phenotypic correlations between the Time 1 predictor variables and anxiety/depression at Time 2, as derived from bivariate Cholesky decomposition, range from 0.32 to 0.41 (Figure 3, Supplementary Table S8). These associations are largely influenced by familial sources (i.e. genetic predisposition and environments that are shared by the twins, such as their home and school environment), which in total account for 61% to 78% of the associations. Analyses are not sufficiently sensitive to significantly distinguish between additive genetic and common environmental influences, excepting the association between Time 1 rumination and anxiety/depression at Time 2, where common environment accounted for 61% of the phenotypic correlation.
Fig. 3. Additive genetic (A), common environmental (C), unique environmental, and familial (A + C) contributions to phenotypic associations between Time 1 (T1) predictors and anxiety/depression symptoms at Time 2 (T2). Nonsignificant estimates are notated ‘ns’. 95% confidence intervals are shown in Supplementary Table S8.
Discussion
Anxiety and depression are complex traits, with adolescence being a core risk period for the emergence of symptoms. Underlying mechanisms may also differ from those in adults (Beesdo et al., Reference Beesdo, Knappe and Pine2009; Hazell, Reference Hazell2021; Hunter & McEwen, Reference Hunter and McEwen2013; Schmaal, Hibar et al., Reference Schmaal, Hibar, Samann, Hall, Baune, Jahanshad, Cheung, van Erp, Bos, Ikram, Vernooij, Niessen, Tiemeier, Hofman, Wittfeld, Grabe, Janowitz, Bulow, Selonke and Veltman2017; Schmaal, Yucel et al., Reference Schmaal, Yucel, Ellis, Vijayakumar, Simmons, Allen and Whittle2017). Notable risk factors for adolescent anxiety/depression include current life stressors, negative thought processes such as rumination, and sleep-related problems (Blake et al., Reference Blake, Trinder and Allen2018; Palmer et al., Reference Palmer, Oosterhoff, Bower, Kaplow and Alfano2018). Here we show that prospectively, perceived stress to everyday events, rumination, and daytime sleepiness account for variation in later anxiety/depression symptoms in young adolescents. However, only perceived stress predicted change in anxiety/depression (i.e. after accounting for anxiety/depression at baseline) although it did not do so independently of rumination and sleepiness, reflecting the considerable overlap among these traits. Familial influences (i.e. additive genetic and/or shared environment) accounted for approximately two-thirds or more of the association between baseline predictor and follow-up anxiety/depression, with shared environment being a prominent component of the association between baseline rumination and follow-up anxiety/depression.
Bidirectional influences were found between all the traits of interest, with the strongest associations over time being for baseline rumination on follow-up perceived stress, and baseline perceived stress on follow-up self-reported symptoms of anxiety/depression. Bidirectional influences involving daytime sleepiness were strongest in the direction of baseline perceived stress on follow-up sleepiness, with baseline sleepiness accounting for a similar amount of variation in follow-up anxiety/depression.
Our results are consistent with studies suggesting that rumination may mediate associations between stressful life events and symptoms of anxiety and depression in adolescents (Hamilton et al., Reference Hamilton, Stange, Abramson and Alloy2015; Hosseinichimeh et al., Reference Hosseinichimeh, Wittenborn, Rick, Jalali and Rahmandad2018; Michl et al., Reference Michl, McLaughlin, Shepherd and Nolen-Hoeksema2013; Skitch & Abela, Reference Skitch and Abela2008). Hosseinichimeh et al. (Reference Hosseinichimeh, Wittenborn, Rick, Jalali and Rahmandad2018), using a system dynamics simulation model approach in 661 young adolescents, determined that rumination contributes to depression by keeping past stressful experiences ‘alive’, which in turn feeds back to stimulate more rumination.
The current study suggests that rumination may also exacerbate responses to everyday stressors (as opposed to keeping stressful life events ‘alive’) — thereby contributing to the maintenance of symptoms of anxiety and depression in adolescents. The daily life stressors scale used in this study reflects, in part, an individual’s perceived control over their environment (Kearney et al., Reference Kearney, Drabman and Beasley1993). Rumination may play a role in enhancing perceptions that environmental pressures are beyond the individual’s capacity to cope, thereby increasing the levels of perceived stress and risk for anxiety and depression.
Daytime sleepiness adds another level of complexity to relationships between perceived stress, rumination, and anxiety/depression. Daytime sleepiness is a common problem among adolescents and is considered to be a clinical marker of adolescent sleep problems (Gradisar et al., Reference Gradisar, Gardner and Dohnt2011). Importantly, adolescence is a vulnerable time for sleep-related problems. Sleep regulatory and other brain systems (e.g. the ‘social’ brain) are still maturing, while psychosocial and societal pressures are increasing (Andrews et al., Reference Andrews, Ahmed and Blakemore2021; Crowley et al., Reference Crowley, Wolfson, Tarokh and Carskadon2018; Kuula et al., Reference Kuula, Pesonen, Merikanto, Gradisar, Lahti, Heinonen, Kajantie and Raikkonen2018). Indeed, these factors are proposed to culminate in a ‘perfect storm’ that may heighten adolescent risk for developing mental health problems (Carskadon, Reference Carskadon2011; Crowley et al., Reference Crowley, Wolfson, Tarokh and Carskadon2018).
Studies specific to daytime sleepiness, especially longitudinal studies in adolescents, are few compared to those exploring other sleep parameters. Daytime sleepiness is an important sleep measure that may provide novel insights in addition to those obtained from standard behavioral sleep parameters (Hong et al., Reference Hong, Choi, Park, Hong, Booth, Joo and Kim2021), although it has been associated with other measures of sleep perception such as sleep quality and insomnia (O’Callaghan et al., Reference O’Callaghan, Couvy-Duchesne, Strike, McMahon, Byrne and Wright2021). Previous longitudinal studies in high-school students have reported reciprocal relationships between daytime sleepiness and both anxiety and depression, as well as perceived study stress predicting later sleepiness (N = 2787, Luo et al., Reference Luo, Zhang, Chen, Lu and Pan2018), and reciprocal associations between sleep disturbances, including daytime sleepiness, and both negative mood and rumination (N = 350, Yip et al., Reference Yip, Xie, Cham and El Sheikh2022). In addition, relatively large cross-sectional studies (N > 1000) have reported associations between daytime sleepiness and perceived psychological stress in high-school students (Chung & Cheung, Reference Chung and Cheung2008; Merdad et al., Reference Merdad, Merdad, Nassif, El-Derwi and Wali2014).
Focusing on anxiety/depression at follow-up as an outcome variable, our analyses indicate that familial influences (i.e. additive genetic and common, or shared, environmental influences) account for most of the phenotypic associations with baseline measures of anxiety/depression, perceived stress, rumination, and daytime sleepiness. But we lack power to disentangle additive genetic and common environmental influences. However, common environment accounts for most of the association between baseline rumination and follow-up anxiety/depression. Common environmental factors are those shared by cotwins, and those associated with rumination may include aspects of family functioning and shared peer group influences and lifestyle factors such as sports participation. For example, parenting styles characterized by high control and protectiveness have been identified as risk factors for the development of ruminative brooding (Manfredi et al., Reference Manfredi, Caselli, Rovetto, Rebecchi, Ruggiero, Sassaroli and Spada2011), while higher levels of brooding are found in adolescents who report having more friends who use alcohol (Hilt, Armstrong et al., Reference Hilt, Armstrong and Essex2017). Further, a study among young adults has posited that exercise may be a buffer against difficulties with emotion regulation, with a less ruminative response style found for those who exercise regularly (Bernstein & McNally, Reference Bernstein and McNally2018).
To estimate genetic and environmental influences for each trait, multivariate Cholesky decomposition of all eight variables (four traits by two timepoints) was conducted. Multivariate analysis has the benefit of increased statistical power to detect effects that are correlated across measures (Schmitz et al., Reference Schmitz, Cherny and Fulker1998). Even so, power was hampered by sample size and confidence intervals were broad, particularly for additive genetic and common environmental estimates. Overall, daytime sleepiness was the most genetic of the traits, with genes accounting for 44% of variance at follow-up. In contrast, we found relatively strong common environmental influences for anxiety/depression at follow-up (accounting for approximately a third of variance) and at both time points for perceived stress and rumination. This likely reflects the correlated nature of common environmental factors on these traits, which would enhance the power of the multivariate analysis to detect them (Schmitz et al., Reference Schmitz, Cherny and Fulker1998).
While our results are broadly consistent with prior work examining anxiety/depression (Nivard et al., Reference Nivard, Dolan, Kendler, Kan, Willemsen, van Beijsterveldt, Lindauer, van Beek, Geels, Bartels, Middeldorp and Boomsma2015; Zheng et al., Reference Zheng, Rijsdijk, Pingault, McMahon and Unger2016) and rumination (Chen & Li, Reference Chen and Li2013; Moore et al., Reference Moore, Salk, Van Hulle, Abramson, Hyde, Lemery-Chalfant and Goldsmith2013), previous adolescent twin studies of anxiety and depression symptoms have reported some inconsistencies, particularly in relation to common environmental sources of influence. For example, in a study of symptoms of anxiety and depression, Nivard et al. (Reference Nivard, Dolan, Kendler, Kan, Willemsen, van Beijsterveldt, Lindauer, van Beek, Geels, Bartels, Middeldorp and Boomsma2015) reported common environmental influences accounting for 11% of variance (37% for additive genetic influences) in Dutch adolescents aged 12 years (N > 1000), but dropping to 0% at age 14 (51% for additive genetic influences), while considerably higher common environmental influences (ranging 20−60%) have been reported for 712 Chinese adolescents aged 10 to 12 years (Zheng et al., Reference Zheng, Rijsdijk, Pingault, McMahon and Unger2016). It is plausible that common environmental influences on anxiety and depression symptoms may be differentially influenced by cultural practices (e.g. Pinquart & Kauser, Reference Pinquart and Kauser2018).
To our knowledge, no twin studies examining the heritability of perceived stress have been specific to adolescents. However, studies in adolescents and/or young adults report higher heritability than found in the current study and no indication of common environmental influence (Michalski et al., Reference Michalski, Demers, Baranger, Barch, Harms, Burgess and Bogdan2017; Rietschel et al., Reference Rietschel, Streit, Zhu, McAloney, Frank, Couvy-Duchesne, Witt, Binz, Consortium, McGrath, Hickie, Hansell, Wright, Gillespie, Forstner, Schulze, Wust, Nothen and Rietschel2017). This is to be expected, as common environmental factors necessarily become less influential as individuals become more independent as they transition to adulthood. Similarly, no independent twin studies of daytime sleepiness specific to adolescents older than 8 years of age were found. Breitenstein et al. (Reference Breitenstein, Doane and Lemery-Chalfant2021) examined daytime sleepiness in children aged 8 and found low heritability (accounting for 27% of variance) and strong common environmental influences (accounting for 66% of variance). In the current study, with older adolescents, common environmental influences were considerably smaller and nonsignificant. Our results are consistent with those found in large adult twin studies, where little or no common environmental influence is found (Desai et al., Reference Desai, Cherkas, Spector and Williams2004; Watson et al., Reference Watson, Goldberg, Arguelles and Buchwald2006).
One possible pathway for genetic and environmental influences on the traits of interest may be through their effects on HPA axis regulation. The HPA axis is a major modulator of responses to external and internal stimuli, including psychological stressors, and alterations in this system have been robustly linked to psychiatric illness, including anxiety and depression (Juruena et al., Reference Juruena, Eror, Cleare and Young2020; Murphy et al., Reference Murphy, Nasa, Cullinane, Raajakesary, Gazzaz, Sooknarine, Haines, Roman, Kelly, O’Neill, Cannon and Roddy2022). Indeed, most depression-related genes identified in replicable gene × environment interactions have been connected to the HPA axis and stress regulation (Gonda et al., Reference Gonda, Petschner, Eszlari, Baksa, Edes, Antal, Juhasz and Bagdy2019; Starr & Huang, Reference Starr and Huang2019). In addition, daytime sleepiness has been associated with a blunting of the cortisol response, which is a marker of HPA axis activity (see review: van Dalfsen & Markus, Reference van Dalfsen and Markus2018). However, associations between rumination, or brooding, and cortisol response have been inconsistent (Hilt, Sladek, et al., Reference Hilt, Sladek, Doane and Stroud2017; Katz et al., Reference Katz, Peckins and Lyon2019; Rnic et al., Reference Rnic, Jopling, Tracy and LeMoult2022; van Santen et al., Reference van Santen, Vreeburg, Van der Does, Spinhoven, Zitman and Penninx2011).
Another mechanism affecting the predictor traits and impacting adolescent mental health may be maturational changes in the circadian timing system during adolescence. Adolescents experience a delayed shift in their sleep onset and offset times that may clash with societal expectations and pressures, leading to daytime sleepiness and reduced capacity to cope with stressors, increased propensity to ruminate, and, ultimately, greater risk of experiencing symptoms of anxiety and depression (Carpenter et al., Reference Carpenter, Crouse, Scott, Naismith, Wilson, Scott, Merikangas and Hickie2021; Carskadon, Reference Carskadon2011; Crowley et al., Reference Crowley, Wolfson, Tarokh and Carskadon2018; Owens et al., Reference Owens, Dearth-Wesley, Lewin, Gioia and Whitaker2016). Consistent with this, circadian clock genes have been posited as a potential nexus for sleep and mood regulation in adolescents (Blake et al., Reference Blake, Trinder and Allen2018; Dueck et al., Reference Dueck, Berger, Wunsch, Thome, Cohrs, Reis and Haessler2015). Further, the HPA axis and mammalian clock gene systems interact, such that stress may regulate clock gene levels (Bolsius et al., Reference Bolsius, Zurbriggen, Kim, Kas, Meerlo, Aton and Havekes2021; Razzoli et al., Reference Razzoli, Karsten, Yoder, Bartolomucci and Engeland2014) and several clock genes are expressed in brain regions implicated in emotion regulation (Kim et al., Reference Kim, Jang, Choe, Chung, Son and Kim2017; Mendoza & Vanotti, Reference Mendoza and Vanotti2019; Patton & Hastings, Reference Patton and Hastings2018).
The current study has some limitations. Sample size limits our ability to disentangle genetic and environmental influences, such that additive genetic and common environmental estimates have broad confidence intervals, with lower bounds generally close to zero. Our measures of wellbeing were obtained from self-report questionnaires, and trajectories of change in adolescents may differ for self-report compared to diagnostic interview (e.g. Long et al., Reference Long, Haraden, Young and Hankin2020). In addition, families with higher socioeconomic status are somewhat over-represented in the QTAB cohort (Strike et al., Reference Strike, Hansell, Chuang, Miller, de Zubicaray, Thompson, McMahon and Wright2022) and results may not best represent individuals in a low socioeconomic environment. Further, this work identifies a relatively small (though important) proportion of factors influencing individual variation in adolescent anxiety/depression symptoms. The problem is a complex one, and diverse approaches are needed to extend our understanding of the underlying issues and to address these problems.
In conclusion, we speculate that HPA axis-related dysfunction and circadian rhythm disturbances (as evidenced by perceived stress, rumination, and daytime sleepiness) may be key drivers in the persistence of anxiety/depression symptoms in young adolescents. However, co-opting these systems in the treatment of psychiatric conditions remains a challenge. Nonetheless, HPA axis genes found to regulate stress responses have been identified as possible drug targets for depression and stress disorders, although many uncertainties remain (Dunlop & Wong, Reference Dunlop and Wong2019; Gonda et al., Reference Gonda, Petschner, Eszlari, Baksa, Edes, Antal, Juhasz and Bagdy2019) and researchers are exploring the potential of plants and phytonutrients to moderate HPA axis activity (Lopresti et al., Reference Lopresti, Smith and Drummond2021). In addition, personalized circadian-targeted therapies for adolescents and young adults with depression and circadian dysregulation have recently been proposed as an enhanced model of care (Crouse et al., Reference Crouse, Carpenter, Song, Hockey, Naismith, Grunstein, Scott, Merikangas, Scott and Hickie2021). Perceived stress, rumination, and daytime sleepiness are easily assessed measures that may hint at underlying HPA axis and circadian rhythm dysregulation and may be valuable markers, in addition to anxiety and depression, for assessing the success of HPA axis and circadian-related interventions. Alternatively, directly addressing perceived stress, rumination, and daytime sleepiness (e.g. Murray et al., Reference Murray, Kurian, Soliday Hong and Andrade2022; Pincus & Friedman, Reference Pincus and Friedman2004; Puolakanaho et al., Reference Puolakanaho, Lappalainen, Lappalainen, Muotka, Hirvonen, Eklund, Ahonen and Kiuru2019; Wassenaar et al., Reference Wassenaar, Wheatley, Beale, Salvan, Meaney, Possee, Atherton, Duda, Dawes and Johansen-Berg2019; Watkins & Roberts, Reference Watkins and Roberts2020) may help to regulate HPA axis and circadian system function and ultimately reduce symptoms of anxiety and depression.
Supplementary Material
To view supplementary material for this article, please visit https://doi.org/10.1017/thg.2022.26.
Acknowledgments
We are grateful to the twins and their families for their willingness to participate in our study. We thank Liza van Eijk, Victoria O’Callaghan, Islay Davies, Ethan Campi, Kimberley Huang, Eleanor Roga, and Michael Day for data acquisition. We acknowledge the Queensland Twin Registry (QTwin) (https://www.qimrberghofer.edu.au/study/queensland-twin-registry-study) for generously sharing database information for recruitment. Recruitment was further facilitated through access to Twins Research Australia, a national resource supported by a Centre of Research Excellence Grant (ID: 1079102) from the NHMRC.
Financial Support
The QTAB project was funded by the National Health and Medical Research Council (NHMRC), Australia (Project Grant ID: 1,078,756 to MJW), the Queensland Brain Institute, University of Queensland, and with the assistance of resources from the Centre for Advanced Imaging and the Queensland Cyber Infrastructure Foundation, University of Queensland.
Conflict of Interest
None.
Ethical Standards
Children’s Health Queensland HREC Reference Number, HREC/16/QRCH/270.
Open access
