Sleep to Internalizing Pathway in Young Adolescents (SIPYA): A proposed model

Abstract

The prevalence of internalizing disorders, i.e., anxiety and depressive disorders, spikes in adolescence and has been increasing amongst adolescents despite the existence of evidence-based treatments, highlighting the need for advancing theories on how internalizing disorders emerge. The current review presents a theoretical model, called the Sleep to Internalizing Pathway in Young Adolescents (SIPYA) Model, to explain how risk factors, namely sleep-related problems (SRPs), are prospectively associated with internalizing disorders in adolescence. Specifically, SRPs during late childhood and early adolescence, around the initiation of pubertal development, contribute to the interruption of intrinsic brain networks dynamics, both within the default mode network and between the default mode network and other networks in the brain. This interruption leaves adolescents vulnerable to repetitive negative thought, such as worry or rumination, which then increases vulnerability to internalizing symptoms and disorders later in adolescence. Sleep-related behaviors are observable, modifiable, low-stigma, and beneficial beyond treating internalizing psychopathology, highlighting the intervention potential associated with understanding the neurodevelopmental impact of SRPs around the transition to adolescence. This review details support for the SIPYA Model, as well as gaps in the literature and future directions.

Introduction

The prevalence of internalizing disorders, namely anxiety and depressive disorders, suddenly increases by an estimated 4% between childhood to adolescence (Ghandour et al., 2019) and has recently been increasing in this age group (Bitsko et al., 2018; Mojtabai, Olfson, & Han, 2016), representing a major public health issue. To help understand this trend, we propose the Sleep to Internalizing Pathway in Young Adolescents (SIPYA) Model, a novel developmental model wherein vulnerability to internalizing disorders in adolescence is exacerbated by sleep-related problems (SRPs) during the transition from late childhood to early adolescence, which may interrupt intrinsic brain networks dynamics and increase internalizing symptoms. Internalizing disorders can be severe, and are associated with significant personal distress, poor academic performance, social withdrawal, as well as suicidal thoughts and behaviors. Importantly, though SRPs may be associated with internalizing symptoms at any point in human development, internalizing disorders that onset before adulthood tend to have higher severity and recurrence than those that onset after adulthood (Fernando et al., 2011; Lim et al., 2013). Additionally, while treatments for adolescent internalizing disorders already exist, the rates of remission remain low (Ginsburg et al., 2011; Kennard et al., 2009). It is therefore critical to invest in prevention efforts, and improve treatments, for adolescent internalizing disorders. Investigating the process through which risk factors, such as SRPs, influence the onset of internalizing disorders in this age range can support novel prevention and intervention strategies.

SRPs in adolescence are prospectively associated with internalizing symptoms (Alvaro, Roberts, & Harris, 2013; Lovato & Gradisar, 2014; McMakin & Alfano, 2015). Common SRPs among children and adolescents include, shortened sleep duration, poor sleep quality, nightmares, daytime sleepiness, wake after sleep onset, increased sleep onset latency, and difficulty sleeping alone (Alfano, Ginsburg, & Kingery, 2007). SRPs that are present during the transition from childhood to adolescence, or peri-puberty, appear to be particularly predictive of internalizing symptoms above and beyond other periods in development (Kelly & El-Sheikh, 2014; Narmandakh, Roest, Jonge, & Oldehinkel, 2020). Sleep is a modifiable behavior and early intervention targeting SRPs during the transition to adolescence may decrease risk for adverse outcomes (Blake, Sheeber, Youssef, Raniti, & Allen, 2017; Clarke et al., 2015; McMakin et al., 2018). Because adolescence is a time of rapid brain and cognitive development, there is interest in how sleep contributes to the development of large-scale brain networks (Dutil et al., 2018; Galván, 2020; Telzer, Goldenberg, Fuligni, Lieberman, & Gálvan, 2015), and, in particular, the default mode network, which has been implicated in internalizing symptoms and disorders (Kaiser, Andrews-Hanna, Wager, & Pizzagalli, 2015; Xu et al., 2019; Zhou et al., 2020). The SIPYA Model, proposes that SRPs during peri-puberty modify the intra- and inter-network dynamics of the default mode network during a crucial time in brain development and thereby increase vulnerability to internalizing disorders later in adolescence (see Figure 1). We discuss 1) support for the connection between adolescent sleep and internalizing symptoms and disorders, 2) support for the connections between the development of the default mode network, sleep, and internalizing psychopathology (i.e., internalizing symptoms, disorders, and related functional impairment), and finally 3) the broader impacts of functional changes in the default mode network on other large-scale brain networks and internalizing psychopathology.

Figure 1:

A timeline on when factors of interest emerge during adolescence: a) Puberty starts a cascade of changes including changes in circadian timing and sleep neurophysiology, which b) interact with ecological constraints, leading to c) an increase in sleep related problems. d) Sleep related problems increase propensity for repetitive negative thoughts, which in turn interrupt sleep. e) The dynamic interplay between these factors increase risk for developing internalizing psychopathology in mid- to late adolescence.

1. Adolescent sleep and internalizing psychopathology

Puberty marks the beginning of a rapid and dynamic stage of development, where endocrinal, neurocognitive, and sociocultural changes, share intertwined and reciprocal relationships (Dahl, Allen, Wilbrecht, & Suleiman, 2018; Goddings, Beltz, Peper, Crone, & Braams, 2019). These pubertal changes include changes in sleep needs, such as delayed bedtimes and increased need for sleep, as well as changes in SRPs when these needs are not met (Laberge et al., 2001; Sadeh, Dahl, Shahar, & Rosenblat-Stein, 2009). Adolescents are generally recommended to obtain 8–10 hours of sleep (Hirshkowitz et al., 2015). Unfortunately, over 50% of adolescents fall short of this recommendation due to a combination of developmental changes, such as a shift in circadian rhythm and a slower build-up of homeostatic sleep pressure such that adolescents across cultures tend to initiate sleep later in the night (Gradisar, Gardner, & Dohnt, 2011; Karan et al., 2021; Owens et al., 2014), and ecological changes, such as earlier school start times, the need to “make up” sleep on weekends, and increased homework (Meltzer, Williamson, & Mindell, 2021).

Shorter sleep duration has a number of cognitive consequences in typically developing adolescents such as impaired attention, increased daydreaming/mind-wandering, daytime sleepiness (Becker et al., 2019) and an increase in driving errors (Garner et al., 2017). Changes in sleep duration can also affect socioemotional functioning. For example, a sleep restriction/extension study with community-recruited adolescents aged 11–15 years, found that sleep restriction, when compared to sleep extension, led to an increase in negative affect as measured through self-report and through pupil response to negative stimuli (McMakin et al., 2016). Additionally, after sleep restriction, adolescents displayed more negative affective behavior with their peers, such as withdrawing from a discussion related to past conflicts. These sleep restriction studies highlight how short-term changes in sleep duration can cause impairments in a range of cognitive and socio-emotional functions.

Given the dramatic short-term effects of shortened sleep duration, it is not surprising that SRPs are closely intertwined with internalizing disorders. In a longitudinal observational study of a large community sample of ninth graders, lower self-reported sleep duration was prospectively associated with higher symptoms of anxiety and depression the next day, and greater intraindividual variation in sleep duration and lower average sleep duration were both uniquely prospectively associated with greater anxiety and depression symptoms (Fuligni & Hardway, 2006). The prospective association between sleep duration and next day internalizing symptoms holds true for actigraphy derived sleep duration as well (Hamann, Rusterholz, Studer, Kaess, & Tarokh, 2019; Kelly & El-Sheikh, 2014). Another study found that actigraphy derived intraindividual variation in bedtimes and sleep onset times were more strongly associated with anxiety symptoms than any average sleep variables (Fletcher et al., 2018).

In addition to SRPs preceding non-clinical increases in internalizing symptoms, SRPs have also been found to precede the diagnosis of internalizing disorders. In a large longitudinal study of community youth, SRPs measured at age 15 years, namely, lower sleep duration, daytime sleepiness, wake after sleep onset, and perception of getting enough sleep, was associated with greater likelihood of receiving a diagnosis of a depression or an anxiety disorder at a later visit (either age 17 or 24 years), even when controlling for internalizing symptoms at baseline (Orchard, Gregory, Gradisar, & Reynolds, 2020). Various SRPs are even included as symptoms of internalizing disorders, such as difficulty falling asleep in generalized anxiety disorder, difficulty sleeping alone in separation anxiety disorder, and decreased or increased sleep duration in depression (American Psychiatric Association, 2013), though the diagnostic criteria does not establish any mechanistic explanation on the relation between SRPs and internalizing psychopathology. Importantly, there is evidence that treatment for internalizing disorders does not always improve SRPs to a clinically significant degree despite clinical impact on other internalizing symptoms (Manglick, Rajaratnam, Taffe, Tonge, & Melvin, 2013; McMakin et al., 2018), and, while SRPs are common among adolescents, not all adolescents with SRPs develop internalizing disorders, emphasizing the need to consider SRPs as separate from, but closely linked to internalizing disorders. Additional common SRPs in adolescent internalizing disorders include bedtime resistance, increased wake after sleep onset, feeling unrested after waking, and decreased subjective sleep quality (for full reviews, see Lovato & Gradisar, 2014; McMakin & Alfano, 2015; Short, Booth, Omar, Ostlundh, & Arora, 2020). As mentioned in these reviews, there is more evidence that SRPs precede the onset of internalizing psychopathology rather than being caused by internalizing psychopathology.

Treatment studies for adolescent internalizing disorders further emphasize the strong link between sleep and internalizing psychopathology. For example, in two different studies examining treatments for adolescent anxiety, better sleep efficiency before treatment (Peterman et al., 2016) and lower parent-reported problems with sleep duration (Wallace et al., 2017), were significantly associated with better treatment outcomes. Another study examining the long term trajectory of depression in children and adolescents found that SRPs were prospectively associated with a recurrence in a depressive episode and increased risk of suicidal thoughts and behaviors at the 12-month follow up (Emslie et al., 2001). In a treatment study for adolescent depression, the addition of an intervention for insomnia symptoms resulted in improved outcomes for depression compared to treatment outcomes for those who received only the depression intervention (Clarke et al., 2015). A similar pilot study examining treatments for children diagnosed with generalized anxiety disorder found that, while adding an intervention targeting SRPs did not significantly improve SRPs and anxiety above and beyond typical anxiety treatment, they did find that SRPs and anxiety improved in both groups, confirming a close relationship with both anxiety and SRPs (Clementi & Alfano, 2020). Finally, in a follow up study for children and adolescents previously diagnosed and treated for anxiety disorders, between-person SRPs, though not within-person fluctuations in SRPs, were associated with higher internalizing symptoms at the next year’s visit and vice-versa (Bai et al., 2020), suggesting that SRPs may be a maintaining factor of internalizing symptoms even in youth previously treated for anxiety. These treatment studies examined SRPs within treatment seeking youth with internalizing disorders, and found SRPs to be related to treatment outcomes, highlighting SRPs as a potential maintaining factor, or barrier to treatment success, and therefore a potentially critical target during treatments for anxiety and depression.

A few studies have also investigated the benefits of targeting SRPs before the onset of internalizing psychopathology. One sleep intervention for adolescents with SRPs and elevated anxiety, though not necessarily clinically elevated anxiety, found that a reduction in SRPs led to a small improvement in anxiety (Blake et al., 2016) and that adolescents with higher levels of internalizing symptoms were more responsive to the sleep intervention than adolescents with lower levels of internalizing symptoms (Blake, Blake, et al., 2018). Another study, that tested an intervention targeting circadian rhythm, found that, in adolescents with excessively late bedtimes, a reduction in eveningness, daytime sleepiness, and parent-reported sleep-wake problems was associated with reduced internalizing symptoms at post-intervention (L. Dong, Gumport, Martinez, & Harvey, 2019). Finally, in a treatment for adolescents diagnosed with primary insomnia, a reduction in insomnia symptoms at the 2 month follow up mediated the effects of treatment on the reduction in internalizing symptoms compared to a waitlist condition, (de Bruin, Bögels, Oort, & Meijer, 2018), though it is important to note that sleep-related anxiety is common in insomnia and that, as constructs, SRPs and internalizing symptoms have some overlap. Overall, studies on targeting SRPs in adolescents shows preliminary evidence that a reduction in SRPs are prospectively associated with internalizing symptoms. However, research is still needed to determine whether a reduction in SRPs in early adolescence lowers risk for developing internalizing disorders in the future and whether there is a sensitive period during which SRPs emerge and are most malleable to alleviating internalizing psychopathology or reducing risk for future problems.

1.1. Why are SRPs in early adolescence particularly important?

Puberty initiates a cascade of physiological changes throughout the body that includes reciprocal changes in sleep physiology and cognition (Lucien, Ortega, & Shaw, 2021; Piekarski et al., 2017). For example, earlier pubertal timing is associated with an earlier circadian shift to later bedtimes (Jessen et al., 2019), implying that many of these changes in sleep are driven by pubertal processes instead of age. Specifically, there is evidence that higher levels of 17-hydroxy-progesterone, a precursor to androgens, in precocious pubertal females is related to later bedtimes, though this study had a relatively small sample size of precocious pubertal females (Jessen et al., 2019). Luteinizing hormone, an early biomarker of puberty that signals the start of changes in gonadal hormones, begins being released exclusively during slow-wave sleep in pubertal children until later in puberty when luteinizing hormone is also released during wake (for review, see: Lucien et al., 2021). Currently little research exists on the effects of sex-hormones, such as estradiol and testosterone, on sleep in humans, as those hormones are hard to detect in pre-pubertal youth (Lucien et al., 2021), but research with animal models, including experimental studies, shows that, in general, changes in sleep-wake cycles are closely related to and likely driven by changes in both male and female gonadal hormones, a finding that is further supported by evidence that sex-differences in sleep tend to emerge around the onset of puberty (Hummer & Lee, 2016; Mong et al., 2011).

Sleep architecture also undergoes significant developmental maturation in adolescence. Sleep architecture refers to the four stages of sleep, as measured through brain activity, which cycle throughout a night of sleep: Rapid Eye Movement (REM) sleep, N1, N2, and N3. The latter three stages are referred to as non-REM sleep, indicated by “N”, and indicate progressively deeper sleep, i.e., more slow wave activity. Slow wave activity, a hallmark of deeper sleep, increases until late childhood and then begins a sharp decline from the beginning of adolescence, reducing by approximately 65% by adulthood (Campbell & Feinberg, 2009; Feinberg & Campbell, 2013). This process of decline in slow wave sleep is closely linked to the onset of puberty, and the increase in luteinizing hormone, whereby earlier puberty is associated with an earlier decline in slow wave sleep, even when controlling for sex (Campbell, Grimm, De Bie, & Feinberg, 2012; Feinberg & Campbell, 2013; Lucien et al., 2021).

These changes in sleep parallel some pubertal and age-related changes in the brain. The decline in slow wave activity follows the same pattern as synaptic pruning, starting from the posterior regions of the cortex to the anterior regions (Feinberg, De Bie, Davis, & Campbell, 2011). Additionally, changes in sleep architecture parallel changes in cortical thickness and gray matter volume, which are thought to change after puberty due to synaptic pruning (Goldstone et al., 2018). These findings suggest that the decline in slow wave activity during sleep, which is associated with the onset of puberty, is likely related to brain maturation in adolescence. However, though these processes parallel each other closely, they need to be examined directly to confirm their relation, and whether SRPs change the trajectory of brain development through changes in sleep physiology.

Within non-REM sleep is the occurrence of another physiological phenomenon in sleep architecture referred to as sleep spindles, which were found to be associated with internalizing symptoms and insomnia in adolescents (for a review, see Trinder, & Allen, 2018). Sleep spindles are short bursts of activity, typically defined as activity in the 12–15 Hz frequency range, that appear during N1 and N2 sleep (Lüthi, 2014). Sleep spindles originate in the thalamus before propagating outwards to the cortex and are associated with a number of cognitive functions including memory and learning (Lüthi, 2014). In a study of 26 children with generalized anxiety disorder and 32 matched controls (aged 7–11 years) who underwent polysomnography (a non-invasive method to measure brain activity during sleep), greater sleep spindle density in the frontal region of the brain was associated with a higher level of worrying regardless of diagnosis (Meers, Ferri, Bruni, & Alfano, 2020). This finding is surprising, considering that other studies have found lower density of sleep spindles in youth with internalizing disorders. One of those studies compared 14 adolescents with social phobia to matched healthy controls (ages 9–17 years) and found widespread reduction in sleep spindles (specifically in the 13–16 Hz range) in the anxious adolescents (Wilhelm, Groch, Preiss, Walitza, & Huber, 2017). Another study similarly found that greater depressive symptoms were associated with lower overall occurrence of sleep spindles (Hamann et al., 2019). Results from a recent longitudinal study, in which polysomnography and clinical measures were collected from 25 children at two timepoints (mean age = 9.52 years, SD = 0.77 at T1, and mean age 16.08 years, SD = 0.9 at T2), may shed some light on this discrepancy. Specifically, results showed that sleep spindles decreased in frontal areas of the brain but increased in other areas such as the central and occipital regions (Bothe, Hahn, Wilhelm, & Hoedlmoser, 2021). Additionally, they found that higher anxious-depressed symptoms at T1 predicted a lower decrease in frontal and central sleep spindles at T2. As polysomnography, which uses electroencephalography, measures brain activity through electrodes outside the skull, precise localization of brain activity is difficult. However, Bothe et al.’s findings reflect a complex relationship between sleep spindles and worry. It is possible that non-typical changes in sleep spindle activity during adolescence may be related to higher internalizing symptoms. More research is needed to understand typical changes in sleep spindles after the onset of puberty before concluding how sleep spindles are related to the development of internalizing psychopathology, but these findings offer preliminary evidence of individual differences that may explain how sleep is related to brain function and internalizing disorders.

Taken together, changes in sleep, triggered by the onset of puberty, seem to parallel brain development during adolescence, and SRPs may leave adolescents at risk for internalizing disorders by interfering with typical brain development. Large-scale longitudinal studies are needed to confirm the relationship between pubertal changes in sleep, SRPs and internalizing disorders. However, the timing of the relationship between SRPs and internalizing disorders, which coincides with puberty-related changes in sleep and the brain, suggest that the adolescent brain may be impacted by puberty-related changes in sleep.

1.2. When are the effects of SRPs on internalizing psychopathology observed?

While SRPs can precede an increase in internalizing symptoms at all stages of development, and intervention studies show that reducing SRPs can reduce internalizing psychopathology at any point in adolescence, the SIPYA Model focuses on late childhood and early adolescence, as longitudinal studies have found SRPs in this age range to be especially predictive of internalizing psychopathology later on. For example, In a sample of 176 community-recruited children (aged 8 years at baseline, then followed up at 10 years, and 13 years), higher self-reported sleep/wake problems were associated with subsequent anxiety and depressive symptoms at all visits while lower sleep duration (as measured by actigraphy) only at age 10 years predicted higher anxiety and depressive symptoms at age 13 years (Kelly & El-Sheikh, 2014). This finding regarding anxiety was replicated in another longitudinal study that followed 2,230 community-recruited children (age at baseline was 11 years) biennially over the course of 15 years (Narmandakh et al., 2020). Results showed that SRPs at only age 11 was prospectively associated with anxiety at age 13 and that SRPs at age 16 predicted anxiety at age 19. Another study that measured SRPs in 199 community-recruited youth at ages 9, 10, 11 and 18 years of age, found that SRPs at age 9 years was significantly associated with higher anxiety and depressive symptoms at age 18 years, above and beyond concurrent SRPs (at age 18 years), which were not significantly associated with anxiety and depression (Shimizu, Zeringue, Erath, Hinnant, & El-Sheikh, 2020). The findings of these studies converge on the particularly strong relationship between SRPs during peri-puberty and future internalizing symptoms.

In addition to the prospective relationship between SRPs and internalizing psychopathology, there is also evidence for a positive feedback loop that leads to further escalations in both SRPs and internalizing symptom severity. For example, in a large sample of 1,573 natural disaster-exposed adolescents followed for 2.5 years (mean age at baseline: 15 years, SD = 1.26), SRPs were prospectively associated with different anxiety symptoms (generalized anxiety, panic, and school phobia), while generalized anxiety symptoms were prospectively associated with later SRPs (F. Geng et al., 2018). A similar pattern was found in the Great Smoky Mountain Study (a longitudinal study of 1,420 community-recruited youth aged 9–16 years) in that, when controlling for age, sex, pubertal status, and comorbidity, SRPs were associated with increased likelihood of later receiving a diagnosis in any anxiety disorders but especially generalized anxiety disorder and, conversely, that a generalized anxiety disorder or depression diagnosis was associated with later increases in difficulty falling asleep, nightmares, and early awakenings (Shanahan, Copeland, Angold, Bondy, & Costello, 2014).

To summarize, while there is generally a strong relationship between SRPs and internalizing psychopathology at all points in development, SRPs during late childhood are particularly associated with internalizing disorders later in adolescence. Furthermore, there is a bidirectional relationship between SRPs and internalizing psychopathology, which raises the possibility of a feedback loop between SRPs and internalizing psychopathology.

1.3. What could explain the relationship between SRPs and internalizing disorders?

The association between SRPs and internalizing symptoms may emerge from worry and rumination, which are core symptoms of anxiety and depression respectively. A broader term for worry and rumination is repetitive negative thinking (RNT), a transdiagnostic construct that refers to passive or uncontrollable and repetitive thinking focused on negative content (Ehring & Watkins, 2008). RNT is found across both anxiety and depressive symptoms among adults (Drost, van der Does, van Hemert, Penninx, & Spinhoven, 2014; Wahl et al., 2019), adolescents, and children (Klemanski, Curtiss, McLaughlin, & Nolen-Hoeksema, 2017; Rood, Roelofs, Bögels, & Alloy, 2010). Among adolescents, RNT is positively associated with both depression and anxiety symptoms (McEvoy et al., 2019). RNT is found to be better at characterizing adolescent anxiety and depressive symptoms than rumination or worry separately (Klemanski et al., 2017) and partially mediates the comorbidity between anxiety and depression (Spinhoven, van Hemert, & Penninx, 2018, 2019). Additionally, RNT seems to mediate the effects of stressors on internalizing symptoms, which means the risk of life-stressors causing internalizing psychopathology is much lower if one does not have RNT after experiencing stressors (Michl, McLaughlin, Shepherd, & Nolen-Hoeksema, 2013). Furthermore, similar to SRPs, increases in RNT precede increases in internalizing symptoms (Hughes et al., 2019; Raes et al., 2020; Spinhoven et al., 2018; Topper, Emmelkamp, Watkins, & Ehring, 2017). Therefore, RNT is likely an important component in the emergence of anxiety disorders and depression. The temporal relationship between SRPs and RNT, however, is more complicated.

Cross-sectionally, there is a consistent positive link between RNT and SRPs such as greater sleep onset latency (Huang et al., 2020), shorter sleep duration (Nota & Coles, 2015), and later bedtimes (Stewart, Gibb, Strauss, & Coles, 2020). However, the temporal direction of this relationship is more complex. One study with undergraduate students found that higher RNT is prospectively associated with higher SRPs 3 weeks later, though they did not test the opposite relationship (Takano, Iijima, & Tanno, 2012). When using an ecological momentary assessment method with undergraduates, another study found that RNT in the evening affected subsequent sleep onset latency, total sleep time, and sleep efficiency (Takano, Sakamoto, & Tanno, 2014). While they did not find that SRPs directly influenced RNT, they did find that reduced sleep efficiency was significantly associated with reduced positive affect in the morning, which was significantly associated with increased RNT in the morning. Because RNT tended to auto-correlate throughout the day, higher RNT in the morning raised the likelihood of RNT before sleep. These findings suggest there is a feedback loop between SRPs and RNT, which the authors propose can be broken by reducing RNT in the evening before sleep. Another recent study monitored daily sleep and RNT in undergraduates over the course of 18 days (Stewart & Coles, 2021). They replicated the bidirectionality of the association between SRPs (specifically shorter sleep duration and later bedtimes) and RNT. When controlling for mood, however, RNT was no longer significantly associated with future SRPs, while SRPs were still associated with future RNT, implying that SRPs may be a stronger predictor of RNT than vice versa. SRPs and RNT are also closely associated in adolescents (Brodar, La Greca, Hysing, & Llabre, 2020; Stewart et al., 2020), though no study has used an ecological momentary assessment method to test this potential feedback loop in adolescents.

Further studies in RNT and SRPs are needed that use an ecological momentary assessment method with a focus on which SRPs tend to be most predictive of RNT, as SRPs are a modifiable behavior. Additionally, while these studies show a potential pathway between SRPs and internalizing symptoms (RNT, specifically), these studies do not shed light on why there is a particularly strong prospective relationship between SRPs in late childhood and internalizing symptoms in adolescence, especially considering that the prevalence of SRPs is higher in older adolescents (Gariepy et al., 2020; Karan et al., 2021). The reason for the increase in SRPs is potentially because early adolescence is a time marked by a mixture of social, psychological, and biological changes. Adolescents may experience increased interest in pursuing peer approval and romantic interests (Forbes & Dahl, 2010), academic pursuits and other activities (e.g. religion, sports, gaming, etc.). Unsurprisingly, adolescents report often feeling worried about dealing with academic load, peer approval, and friendships (Bailey, Giles, & Rogers, 2015) including at night (Hiller, Lovato, Gradisar, Oliver, & Slater, 2014). While these factors are likely more salient as adolescents get older, in early adolescence, this socio-cultural context is further complicated by the complex physiological changes associated with the onset of puberty, such as changes in cognition and sleep needs as well as changes in brain development.

2. The Default Mode Network

Internal mentation, which includes RNT, implicates a network of brain regions called the default mode network. The default mode network is a large-scale brain network that was first identified as the network that is consistently active when the brain is not engaged in a task, as measured through resting-state functional MRI (fMRI; Raichle et al., 2001; Shulman et al., 1997). Subsequent task-based studies clarified that regions of the default mode network do increase in activity during goal directed behaviors if the fMRI task requires functions relevant to the default mode network (Andrews-Hanna, Reidler, Sepulcre, Poulin, & Buckner, 2010; Laird et al., 2009; Spreng, Stevens, Chamberlain, Gilmore, & Schacter, 2010). Broadly, the function of the default mode network involves internal mentation, which includes mind-wandering/day-dreaming (Mason et al., 2007; Philippi et al., 2021; Smallwood, Brown, Baird, & Schooler, 2012), social cognition (Mars et al., 2012; Schilbach, Eickhoff, Rotarska-Jagiela, Fink, & Vogeley, 2008), and self-referential processes, such as remembering the past, (Andrews-Hanna, Reidler, Huang, & Buckner, 2010; Buckner, Andrews-Hanna, & Schacter, 2008) or thinking about the future (Karapanagiotidis, Bernhardt, Jefferies, & Smallwood, 2017). It is important to note that resting-state fMRI may reveal spontaneous or intrinsic network activity offering complementary functional information compared to task-based fMRI, whereby research participants complete a task designed to recruit a specific cognitive function (Lv et al., 2018; Smitha et al., 2017). It is also important to note that, while activity in a brain region may be associated with a particular task-related function, the reverse cannot necessarily be inferred, meaning activity in a region of interest cannot be assumed to signify the presence of a particular cognitive function (Poldrack, 2011).

Though the exact anatomy of the default mode network is still debated, the most consistent regions include the ventromedial prefrontal cortex (vmPFC), dorsomedial prefrontal cortex (dmPFC), precuneus, and posterior cingulate cortex (PCC; Andrews-Hanna, 2012; Andrews-Hanna, Reidler, Sepulcre, et al., 2010; Andrews-Hanna, Smallwood, & Spreng, 2014; Laird et al., 2009; Raichle, 2015). The default mode network is functionally integrated within its own local regions, functionally segregated from other networks, and retains a functional connection between the anterior (e.g., mPFC) and posterior (e.g., PCC and precuneus) regions (Fair et al., 2008), forming distinct hubs and subsystems (Andrews-Hanna, Reidler, Sepulcre, et al., 2010; Laird et al., 2009). One breakdown proposed by Andrews-Hanna et. al, (2010) divides the default mode network into a core network and two subsystems. The first subsystem, dubbed as the dmPFC subsystem, contains the dmPFC, temporoparietal junction, lateral temporal cortex, and temporal pole. The second subsystem, dubbed as the medial temporal lobe (MTL) subsystem, contains the vmPFC, along with the hippocampal formation, parahippocampal cortex, retrosplenial cortex, and posterior inferior parietal lobule. Finally, the midline core of the default mode network contains the PCC, precuneus, and anterior mPFC.

2.1. The development of the default mode network

In general, major brain networks, such as the default mode network, are largely present by age 2 years but continue to be refined through late childhood and adolescence (Grayson & Fair, 2017). When examining adolescent brain development by age cross-sectionally, resting-state functional connectivity and strength within the default mode network and between the default mode network and other networks increased linearly with age (Gu et al., 2015; Solé-Padullés et al., 2016). A more recent longitudinal study, with 305 children between ages 6–12 years found similar results in that overall resting-state connectivity strength and connectivity pattern increased linearly with age (Fan et al., 2021). Additionally, compared to an adult group, default mode network connectivity strength and efficiency in the child group was lower at all ages, suggesting that the default mode network is still developing into late adolescence. However, the linear trajectory and lack of sex differences does not imply an important role of pubertal onset in the development of the default mode network. A possible reason for this is that these studies used age as a measure of development instead of pubertal stage, and many developmental changes are more correlated with pubertal changes than age (Goddings et al., 2019; van Duijvenvoorde, Westhoff, de Vos, Wierenga, & Crone, 2019).

In early adolescence, pubertal development is an important consideration in brain development. In a longitudinal study, completed at age 10 years and again at age 13 years, involving a fMRI task where participants evaluated either social traits or academic traits in oneself or another, higher pubertal stage at the follow-up scan (while controlling for baseline pubertal stage) correlated with increased activation in the vmPFC during an fMRI task involving self-related social evaluations (Pfeifer et al., 2013). This study did not find any sex-related differences in pubertal development, however, possibly due to the small sample and uneven male to female ratio, which may have obscured any sex differences. Another study specifically looked at dmPFC activation in 35 female youth (ages 11–13 years) and found that, when completing 14 minutes of an fMRI task designed to evoke either social emotions (e.g., embarrassment) or general emotions (e.g., disgust), advanced pubertal development was related to greater inter-network connectivity between the dmPFC and other brain regions previously associated with social information processing (Klapwijk et al., 2013). While this study is cross-sectional and had a small sample, the results were generally consistent across different measures of pubertal development (physician rating, self-report, and hormonal assay). A more recent longitudinal study (n=98, ages 6–18 years) found that resting-state functional network organization (e.g., efficiency) of the default mode network, suddenly begins to increase shortly after the first signs of self-reported pubertal development (Gracia-Tabuenca, Moreno, Barrios, & Alcauter, 2021). As female youth in this study tended to show the first signs of pubertal development about a year before same-aged male youth, default mode network development in females reached this sudden increase in network organization at an earlier age than male participants. While this study has many strengths, such as a large sample size and longitudinal design, the authors acknowledged a number of weaknesses, such as the short 5-minute resting-state protocol, the use of self-report to measure pubertal stage, and only some participants followed up with one or two more scans. Because of these weaknesses, further replication with more recent resting-state protocols, a variety of pubertal measures starting from before age 10 years, and more follow up scans across the full sample are needed to confirm this non-linear trend. Finally, a study with 304 14-year old adolescents found significant sex by puberty interactions in that the default mode network resting-state functional connectivity decreases in female adolescents over pubertal development while increasing in male adolescents (Ernst et al., 2019). This study is strengthened by the narrow age range, effectively removing age as a factor, but fails to capture early pubertal development.

A significant limitation to this work is that many findings from fMRI are confounded by motion during image acquisition in all sample demographics but especially in younger samples. Younger research participants tend to exhibit more movement during image acquisition than older participants, requiring motion correction techniques to accurately interpret age effects on functional connectivity above and beyond motion (Grayson & Fair, 2017; Power, Schlaggar, & Petersen, 2015). Motion correction techniques and image quality practices are still being standardized and therefore many fMRI results must be interpreted, or re-interpreted, in light of these emerging issues (Maknojia, Churchill, Schweizer, & Graham, 2019; Power et al., 2015).

Methodology for studying pubertal development has similarly evolved over time to account for an emerging understanding of this complex developmental period. Pubertal development is complicated by a number of factors, as reviewed by Dai and Scherf (2019), including the high correlation between age and pubertal status, difficulty with measuring hormones which differ by sex and fluctuate over time, non-linear developmental course, need for multimodal measures, etc. For studies on pubertal neurodevelopment, these factors are important to consider in addition to factors related to motion in this age range. While the study by Gracia-Tabuenca et al. (2021), suggests that pubertal development is strongly linked to functional brain development, further research is needed to elucidate which pubertal processes are related to which brain networks and how. Future research would benefit from longitudinal design and multimodal measures of puberty that can capture the cyclical nature of hormones, to help reach causal conclusions in the effects of puberty on brain development.

2.2. Sleep and the default mode network

In addition to changes in functional connectivity over the course of development, the default mode network and other regions also undergo functional changes due to differences in sleep. In general, cognitive deficits, such as difficulty concentrating, increase after a night of sleep deprivation or a night of poor quality sleep (Krause et al., 2017). Total sleep deprivation is associated with lower functional connectivity between the mPFC and precuneus during resting state fMRI, which was associated with lower vigilance during a vigilance task (Chen et al., 2018), though these findings need to be confirmed with a follow-up study that more rigorously controls for potential motion artifacts resulting from 24-hour sleep deprivation.

Insomnia is associated with increased resting-state functional connectivity within the left dmPFC and lower resting-state functional connectivity within the PCC and precuneus compared to healthy adults, demonstrating an imbalance in resting-state activity within the default mode network between the dmPFC subsystem and the midline core of the default mode network (Yu et al., 2018). The dmPFC subsystem and midline core regions of the default mode network have previously been broadly implicated in self-referential processing and autobiographical recollection respectively (Andrews-Hanna, Reidler, Sepulcre, et al., 2010). Interestingly, these insomnia-related differences in resting-state functional connectivity of the default mode network map onto common symptoms of depression, namely excessive rumination about the self and reduced accuracy in autobiographical recollection (Lyubomirsky, Caldwell, & Nolen-Hoeksema, 1998). Insomnia has also been associated with lower resting-state functional connectivity between the anterior and posterior regions of the default mode network (X. Dong et al., 2018). Conversely, another study did not find differences in default mode network connectivity between insomnia patients and healthy controls, but did find that greater resting-state functional connectivity between hippocampal regions and default mode network regions was associated with worse sleep the night before in both insomnia patients and healthy controls (Regen et al., 2016).

Effects of SRPs on the default mode network connectivity are also found in adolescents. In a within-person design study of 18 adolescents aged 13–15 years, a night of sleep restriction (4 hours), compared to a night of normal sleep, preceded weaker local resting-state connectivity within the PCC (Robinson, Erath, Kana, & El-Sheikh, 2018). Similar results were found in an observational study examining sleep patterns in 45 adolescents aged 14–18 years (Tashjian, Goldenberg, Monti, & Galván, 2018). Specifically, the authors found that lower sleep continuity (actigraphy measured sleep efficiency, number of night awakenings, and duration of night awakenings) was associated with weaker resting state functional connectivity between the posterior region (PCC and precuneus) of the default mode network with other areas of the default mode network (i.e., decreased intra-network connectivity). With regards to associations with brain structure, shorter sleep duration, later sleep timings, and lower sleep continuity was correlated with cortical thinning in major regions of the default mode network (e.g., PCC) in early to mid-adolescent participants but not older adolescents and young adults (Jalbrzikowski et al., 2021), highlighting the potential sensitivity of this developmental period to sleep. Similar to the study by Yu et al., (2018), these studies with adolescents found a decrease in resting-state functional connectivity within the PCC, a region of the midline core and major network component involved in integrating other regions of the default mode network as elaborated in Andrews-Hanna, Reidler, Sepulcre, et al., (2010).

Research on SRPs and functional connectivity during wake have a number of potential confounds, such as severity of SRPs, chronic versus acute SRPs, and the interaction between the timing of the fMRI scan and an individual’s unique circadian rhythm, all of which require precise characterization of an individual’s sleep habits. Even without precise characterization of SRPs, this cursory overview highlights how SRPs can affect functional network connectivity and coordination within the default mode network. There is still need to investigate whether the SRP-induced change in default mode network activity leads to interruptions from off-task thinking (i.e., mind-wandering) and which factors contribute to negative content in off-task thinking. Additionally, while studies with insomnia patients revealed similar results as studies with short term sleep deprivation, more longitudinal studies are needed to investigate how SRPs and default mode network function influence each other over time, and whether those changes are different if SRPs are present during peri-puberty compared to other periods of development. Finally, sleep spindles were found to communicate between regions of the default mode network during sleep (Fang et al., 2020), but so far, no study has examined the link between sleep spindle communication across regions of the default mode network and subsequent cognitive function. Because sleep spindle activity undergoes many changes during adolescence (Goldstone et al., 2019), and sleep spindle activity is related to internalizing symptoms (Bothe et al., 2021), it is important to examine how sleep spindle activity in the default mode network changes during adolescence and whether those changes predict, or are predicted by, internalizing symptoms such as RNT.

2.3. Internalizing psychopathology and the default mode network

The role of the default mode network in internalizing disorders has been a subject of intense investigation, as core features of these disorders are related to internal mentation, specifically RNT (i.e., ruminations about the past and/or worries about the future). In a large meta-analysis, the frontal region of the default mode network, and in particular greater fMRI task-evoked activity in the dmPFC, is heavily implicated in higher levels of fMRI task-evoked rumination (Zhou et al., 2020), suggesting that a decrease in functional connectivity in this region may signal a decrease in internalizing symptoms. Indeed, patients with remitted depression showed lower resting-state functional connectivity in the frontal regions of the default mode network compared to patients with current depression and, interestingly, control participants (never diagnosed with depression) as well (Q. Dai et al., 2018). The authors propose that the MDD treatment may involve an overcompensation in suppressing rumination and that the lack of equivalence in default mode network function between remitted MDD and controls may signal a trait-like difference between adults at risk for MDD and those not at risk. Regarding the current MDD patients’ increased activity in the frontal default mode network, another study similarly found that patients with MDD, when compared to controls, have less deactivation (higher task-evoked activity) in the mPFC when passively viewing or actively reappraising negative pictures, potentially reflecting how self-referential thoughts among individuals with MDD can interfere when processing external negative stimuli (Sheline et al., 2009).

The increase in activity in the dmPFC with depressive symptoms was not found in adolescents, however. With regards to adolescents and puberty, one study compared 20 adolescents with major depressive disorder to 29 non-depressed adolescents (ages 10–18 years) and found that, when completing a task involving neutral versus social words, depressed adolescents showed lower activation in the mPFC when viewing social words (Silk et al., 2017). While they found that task-evoked activity in the mPFC increased with age, once they added self-reported pubertal status, age was no longer significant. Because of the age range, pubertal status and age were likely highly correlated, making it difficult to differentiate between age and puberty effects. A larger sample with a more restricted age range may help to parse age effects from puberty. With regards to puberty, advanced pubertal status was related to lower resting-state functional connectivity within the default mode network in females, and the opposite in males (Ernst et al., 2019). In an exploratory analysis, the authors found that lower resting-state functional connectivity in the dmPFC at age 14 years is negatively associated with pubertal status and prospectively associated with higher internalizing symptoms at age 16 years. This finding suggests that puberty-related changes in resting-state functional connectivity in the default mode network may be related to increased vulnerability in female adolescents, or increased resilience in male adolescents. However, these findings are in direct contradiction with findings from the meta-analysis by Zhou et al., (2020). It is difficult to draw any conclusions on this discrepancy. It is important to design, and allocate resources for, longitudinal research examining the effects of puberty-related constructs (e.g. timing of onset, pace of pubertal development, pubertal status, specific hormonal changes) on brain development as well as the evolution of internalizing psychopathology over development.

Social cognition is also highly relevant to RNT, such as in the case of social anxiety, and is related to activations in the default mode network. Results from studies with social phobia patients found that the PCC was associated with higher activation compared to controls during fMRI tasks (Gentili et al., 2009; Maresh, Allen, & Coan, 2014) and that overall, there is decreased resting-state functional connectivity between the posterior and anterior default mode network, with increased spontaneous activity in the PCC during resting-state fMRI (Yuan et al., 2018), though a different study with a smaller sample size found an increase in the resting-state functional connectivity between the anterior and posterior default mode network in social phobia (Rabany et al., 2017). The PCC has previously been implicated in social cognition. One experiment found that the PCC has greater activity when participants read about a character’s thoughts but not when they read other socially relevant information, such as a character’s appearance or bodily sensations during an fMRI task (Saxe & Powell, 2006). This finding contrasts with the effects of SRPs on PCC activity. This may be due to differences in content of RNT between different internalizing disorders. For example, in social phobia, RNT might be focused more on what others think, which is a diagnostic criteria (American Psychiatric Association, 2013), while, at least in some cases, suicidal thoughts and behaviors can stem from a feeling of disconnect from others, as proposed by the interpersonal theory of suicide (Van Orden et al., 2010). In a cross-sectional study on suicidal thoughts and behaviors in adolescents diagnosed with depression, higher suicidality was associated with lower resting-state functional connectivity between the PCC and other brain regions and higher resting-state functional connectivity between precuneus and other brain regions (Schreiner, Klimes-Dougan, & Cullen, 2019), which partially maps onto findings on the effects of SRPs on default mode network activity, namely a decrease in resting-state PCC activity and connectivity (Robinson et al., 2018; Yu et al., 2018). The link between SRPs and suicidality is not surprising as suicidal thoughts and behaviors are known to be exacerbated by insomnia and other SRPs (Blake & Allen, 2020; Goldstein & Franzen, 2020). As demonstrated, the default mode network is critical in understanding internalizing psychopathology, but internalizing psychopathology, and the brain, are complex and dynamic and therefore not easily characterized into a simple mechanism.

3. Internalizing psychopathology through a triple network model framework

Beyond the default mode network, interactions between large-scale brain networks (i.e., inter-network dynamics) also play an important role in internalizing psychopathology (Bagherzadeh-Azbari et al., 2019; H. Geng, Li, Chen, Li, & Gu, 2016; Rabany et al., 2017). The default mode network is one of the three main components of the triple network model, a theoretical framework proposed by Menon (2011), which posits that the default mode network, along with the central executive network and salience network coordinate higher cognition, and that imbalance within or between these networks can help explain psychopathological impairment (Menon, 2011). The triple network model is a helpful framework for conceptualizing and interpreting large-scale brain network interactions and, in the case of the SIPYA Model, can help interpret the larger-scale impact of changes in the default mode network in relation to SRPs and internalizing psychopathology. According to the triple network model, the central executive network is active during goal oriented, often outward, tasks involving manipulating information in working memory, problem solving, and decision making. The central executive network includes the dorsolateral prefrontal cortex (dlPFC) and posterior parietal cortex (PPC; Menon, 2011). The salience network broadly monitors the body and environment for salient information and then modulates between the default mode network and central executive network as needed. The salience network includes the dorsal anterior cingulate cortex (dACC), the anterior insula (AI) and amygdala (Menon, 2011).

With regards to anxiety and the triple network model, a meta-analysis of 23 resting-state fMRI studies, in which 466 young adults diagnosed with anxiety disorders were compared with 508 healthy controls, found that anxiety was associated with lower resting-state functional connectivity between the salience network and default mode network and lower resting-state functional connectivity between the salience network and central executive network, which may manifest as difficulty with down-regulating self-referential processes in the presence of motivationally salient environmental stimuli (Xu et al., 2019). Additionally, the metanalysis revealed lower resting-state functional connectivity (i.e., lower negative correlation/anti-correlation) between the default mode network and central executive network, which supports the idea that anxious individuals are often unable to down regulate attention to themselves while pursuing goal-directed activities. This finding was partially replicated with adolescents, in that higher trait anxiety was associated with lower resting-state functional connectivity between regions of the salience network and default mode network (H. Geng et al., 2016).

With regards to depression, a metanalysis of 25 resting-state fMRI studies, in which 556 adults and mid- to late adolescents diagnosed with depression were compared to healthy controls found lower resting-state functional connectivity within the central executive network (PPC and dlPFC), increased functional connectivity within the default mode network, and increased functional connectivity between the default mode network and dlPFC but not PPC (Kaiser et al., 2015). The relationship between the default mode network and central executive network found in depression is similar to the relationship found in anxiety, in that there is a positive correlation in resting-state activity between the dlPFC and the default mode network while in anxiety there is decreased anti-correlation. In both cases, the default mode network and regions of the central executive network are imbalanced but not identically so. This difference may be due to the heterogeneity of internalizing symptoms. For example, with regards to depression, one resting-state fMRI study with adolescents found that depression severity is related to higher resting-state functional connectivity between the dmPFC and frontal regions of the central executive network while severity of anhedonia (a common but not required symptom of depression) was associated with higher resting-state functional connectivity between the ACC and dmPFC (Rzepa & McCabe, 2018). For another example, state anxiety has different resting-state functional connectivity correlates than trait anxiety (Dennis, Gotlib, Thompson, & Thomason, 2011; H. Geng et al., 2016). Further research on the triple network model and internalizing psychopathology likely requires a more precise and sophisticated conceptualization of the different types of symptoms, such as types of RNT (focus on self vs. others) or differentiating between distress symptoms and fear/avoidance symptoms.

SRPs are also related to alterations in functional connectivity of the triple network model. The most commonly studied SRP is short sleep duration where individuals experience decreased anti-correlation (implying imbalance) between the activity in the default mode sand central executive networks both during attention tasks and during resting state scans after a period of sleep deprivation (Cai, Mai, Li, Zhou, & Ma, 2021; C. Dai et al., 2020; De Havas, Parimal, Soon, & Chee, 2012; Gujar, Yoo, Hu, & Walker, 2010; Kaufmann et al., 2016). Behaviorally, lower anti-correlation between changes in activity in regions of the default mode network and central executive network is associated with instances of mind-wandering as well as increased errors when completing an fMRI task (Christoff, Gordon, Smallwood, Smith, & Schooler, 2009). Specifically, during task-unrelated thoughts (e.g., mind-wandering or daydreaming) the default mode network increases in activation but also recruits regions of the central executive network. As SRPs also predict lower anti-correlation between these regions, it highlights a potential mechanism through which SRPs increase the propensity for RNT. Conversely, lower anti-correlation between the default mode network and central executive network resting-state functional connectivity seems to signal vulnerability to sleep deprivation, meaning sleep deprivation will be more detrimental for cognitive functioning in those with lower intrinsic anti-correlation between the default mode network and central executive network (Yeo, Tandi, & Chee, 2015), potentially causing a cycle of cognitive difficulty. This cycle is similar to the RNT-SRP feedback loop, whereby RNT at night predicts low mood and RNT the next morning, which may be propagated throughout the day until bedtime, starting the cycle over again (Takano et al., 2014). Further longitudinal studies would be necessary to determine whether the RNT-SRP feedback look has any correlates in the Triple Network Model framework.

Though there are not many studies on the influence of puberty on the Triple Network Model yet, a few studies hint that there may be some sex differences that arise in the development of these three large-scale brain networks. One study found that, when completing a self-referential task, female adolescents had stronger functional connectivity between the dlPFC (central executive network) and the precuneus/mPFC ( default mode network) than male adolescents (Alarcón, Pfeifer, Fair, & Nagel, 2018). This study did not find any differences related to pubertal stage, though that may be due to the older age range, at which point many adolescents are post-pubertal. Another study with 112 peri-pubertal adolescents did not find any differences in levels of rumination or functional connectivity, between male and female adolescents, but did find that higher resting-state functional connectivity within the salience network was associated with higher levels of rumination in girls but not boys (Ordaz et al., 2017). This study constricted the sample to early puberty, with many girls and boys showing no physical signs of pubertal maturation, to reduce any confounding effects of pubertal development on sex differences, so these findings may change as adolescents enter the later stages of puberty. There is additionally evidence in the adult literature of a close relationship between fluctuations in sex hormones and brain network connectivity and the potentially lasting effects of pubertal hormones on brain development. For example, oral contraceptive use, when initiated around pubertal onset, is related to higher-resting state functional connectivity within the salience network, which suggests that hormone levels during puberty likely have long term effects on brain network connectivity (Sharma, Fang, Smith, & Ismail, 2020). Additionally, the effect of fluctuations in ovarian hormones during the menstrual cycle on negative mood in adults may be mediated by its effects on the default mode network and salience network resting-state functional connectivity (Andreano, Touroutoglou, Dickerson, & Barrett, 2018). Research on changes in pubertal hormones and functional brain development in adolescents so far has found mixed results, largely due to the complexity involved in understanding and measuring hormonal fluctuations during puberty (J. Dai & Scherf, 2019), but these adult studies suggest that pubertal hormones are important to consider in understanding brain development and vulnerability to internalizing psychopathology. Overall, these studies suggest that puberty is likely an important factor in understanding the development of the three networks as there are few sex differences in internalizing symptoms and functional connectivity before puberty, but after puberty, sex differences are more apparent.

A few studies have examined how sleep-related changes in functional connectivity relate to changes in internalizing symptoms. In one study, with 18 healthy adults, 24 hours of total sleep deprivation was related to subsequent higher state-anxiety and increased task-evoked activity in regions of the salience network (amygdala and dACC), as well as the default mode network (specifically the mPFC) when viewing emotional vs. neutral stimuli (Ben Simon, Rossi, Harvey, & Walker, 2020). In another study comparing depressed patients with insomnia and depressed patients without insomnia, insomnia symptoms were associated with increased resting-state activity in regions of the salience network but did not find between network differences or, interestingly, differences in resting-state activation within the default mode network (Liu et al., 2018). With regards to adolescents and internalizing symptoms, effects of adolescent sleep continuity and default mode network resting-state functional connectivity overlapped with critical regions related to suicidal thoughts and behaviors in adolescents with major depression in that greater suicidal thoughts and behaviors and lower sleep quality are both related to weaker resting-state functional connectivity between the PCC and other regions (Schreiner et al., 2019; Tashjian et al., 2018) though this relationship has not been tested in a single sample. As of yet, no studies have examined the effects of SRPs on internalizing symptoms through a Triple Network Model framework. Further research is needed to look at these relationships within a single early adolescent clinical sample.

Summary and Limitations

This review describes a possible explanation to the increase in the prevalence of internalizing disorders in adolescence by proposing the SIPYA Model, in which SRPs during peri-puberty increase vulnerability to internalizing disorders by disrupting the inter- and intra-default mode network functional connectivity during a major developmental period, thereby increasing the propensity for RNT. In this paper, we have established a number of developmentally sensitive processes that support this model: 1) puberty-related changes in sleep physiology interact with an increase in SRPs in late childhood, 2) in general, and particularly during the transition to adolescence, SRPs tend to precede internalizing symptoms, such as RNT, 3) the default mode network undergoes many changes after the onset of puberty, and is heavily implicated in internalizing symptoms, including RNT, 4) SRPs and internalizing symptoms, including RNT, are also related to imbalance between the default mode network and the central executive and salience networks.

These findings, when organized into the SIPYA Model, establish the transition to adolescence as a sensitive period for when the effects of SRPs on functional brain development and socio-emotional well-being are most apparent, and emphasize the importance of addressing those SRPs to prevent and treat internalizing disorders. Previous research has already shown that some SRPs, such as sleep onset latency and sleep duration, are modifiable in anxious and depressed youth (Clarke et al., 2015; McMakin et al., 2018). However, treatments and interventions targeting sleep can benefit from more precise understanding of which dimensions of sleep in particular affect brain function and related cognitive functions, as well as when and how these effects occur.

The relationships between different developmental factors are not straightforward and the SIPYA Model may only account for one possible pathway. Early puberty is a short but dynamic time in development, and findings from outside of this window may not apply. For example, sleep spindles change drastically during puberty, but individual changes in sleep spindles may be more informative than simply comparing sleep spindle power/amount/number between adolescents with internalizing disorders and healthy controls. Brain development and function is likely similar in that redundancy, adaptations, plasticity and feedback loops, which differ between individuals, can obscure causal relationships. This is especially the case in research on puberty-related changes in the brain, which adds the complexity of changing hormone cycles (J. Dai & Scherf, 2019). The onset of puberty is caused by the convergence of many factors and is influenced by many biological, psychosocial, and contextual factors, and therefore requires specific measures of puberty (e.g., hormonal assays instead of self-reported pubertal staging) and strong theory driven hypotheses that can account for context.

Similarly, brain activity is fast and dynamic while functional connectivity, as measured through fMRI, is slow and liable to many influences beyond anatomical structure, that may introduce confounds, such as state-dependent mood or head motion (Buckner, Krienen, & Yeo, 2013). Furthermore, fMRI research largely correlates brain activity with an operationalized behavior of interest. While it is possible to determine which regions are more active during a specific task, the reverse, i.e., determining which cognitive process is active given activity in a region of interest, cannot be accurately concluded as there is no evidence to limit brain function in any given region (also known as reverse inference, see: Poldrack, 2011). It is difficult to pinpoint how and why spontaneous activity occurs, such as in the case of resting-state fMRI, and without a stronger theoretical basis, it may be difficult to draw consistent conclusions. Hence, many studies on anxiety and depression find slightly different relationships between functional connectivity within and between major neural networks.

Despite the difficulty in elucidating specific relationships between SRPs, default mode network function, and internalizing symptoms, a few findings seem to be consistent: 1) puberty is associated with dynamic developmental changes in cognitive function and sleep physiology which parallels, and is prospectively associated with, brain development and function, and 2) adolescents experience many SRPs, which can cause emotional and cognitive dysfunction and predict later internalizing disorders. There are currently no longitudinal studies of puberty, sleep, and brain development that utilize both objective and subjective measures to map how these factors interact over time, where they go awry, and when intervention efforts would be most effective. The SIPYA Model offers a framework to help guide future research in investigating these gaps.

Future Directions

The SIPYA Model posits that SRPs are an early and modifiable risk factor for internalizing disorders, giving a window of opportunity to have a positive impact on adolescent development. If supported, this model emphasizes the importance of prioritizing sleep during the transition to adolescence and promotes policy changes that reduce ecological interruptions to sleep (e.g. early school start times, late extracurriculars). Additionally, the effects of SRPs on vulnerability to RNT, and the potential SRP-RNT feedback loop may require modifying treatments to reduce propensity for RNT through sleep interventions or a focus on reducing RNT before bedtime, regardless of developmental stage. Or, conversely, if SRPs during peri-puberty influence brain development, sleep interventions may be most effective during that time, and during other periods of development, different strategies could be prioritized. Furthermore, SRPs are a low-stigma target for interventions with benefits beyond treating internalizing psychopathology, which can increase the reach of interventions to those who may otherwise not seek services. Future research on the possibility of a RNT-SRP feedback loop can further investigate which part of the process are most amenable to modification. In general, sequencing brief interventions to target symptoms that are most modifiable during any given developmental period may support more rapid treatment gains; this is an important consideration given the overloaded mental health system (Vigo, Kestel, Pendakur, Thornicroft, & Atun, 2019), the high rate of treatment dropout among youth who seek clinical services (de Haan, Boon, de Jong, Hoeve, & Vermeiren, 2013) and the increasing emphasis of intervention delivery in primary care and community settings to enhance reach (Fazel et al., 2021).