Circadian, Reward, and Emotion Systems in Teens prospective longitudinal study: protocol overview of an integrative reward-circadian rhythm model of first onset of bipolar spectrum disorder in adolescence

Biological processes that repeat about every 24 h and persist with the same period in the absence of external cues are defined as circadian rhythms [10]. The suprachiasmatic nucleus (SCN), or “biological clock”, regulates circadian rhythms [10, 11], and is highly sensitive to exogenous cues [12]. Circadian rhythms normally are entrained or phase aligned to the physical environment predominately by the external zeitgeber (“time giver”), light [12, 13], but also by nonphotic cues, including daily schedules or social rhythms [14, 15]. Circadian models of BSDs propose that circadian dysregulation leads to bipolar hypomanic and manic (hereafter referred to as hypo/manic) and depressive symptoms and episodes and provides vulnerability to BSD onset [16–19]. Many individuals with BSDs have circadian rhythm alterations of sleep-wake, activity, melatonin, and other hormones even in euthymic periods. Individuals with hypo/manic symptoms also show activity rhythm irregularities as measured by actigraphy and self-report. In addition, individuals with BSDs exhibit clock gene variants that show irregular rhythm patterns [16, 20]. However, not all studies uniformly support a role for circadian dysregulation in BSDs [21–23]. And, an important limitation of prior research is that it has never tested whether circadian rhythm dysregulation provides vulnerability to first onset of BSD prospectively.

According to the social zeitgeber theory [17, 18, 24], life events that disrupt daily social rhythms or schedules precipitate bipolar symptoms by disturbing circadian rhythms (green boxes in Fig. 1). Social rhythm disruption (SRD) events may lead to changes in light exposure, which can phase shift melatonin and other circadian rhythms [12, 13, 25], or may disrupt circadian rhythms through their effects on non-photic zeitgebers [17, 18, 26]. Indeed, individuals with BSDs exhibit less social rhythm regularity than controls [27–29], and this irregularity predicts greater mood symptom variability [30] and shorter time to recurrence of bipolar depressive and hypo/manic episodes [29]. The efficacy of treatments involving social rhythm stabilization for patients with BSDs also supports the role of SRD in bipolar disorder [31, 32]. In people with BSD, SRD events were more likely to occur prior to manic or depressive episodes relative to control periods [33–35], and an increase in SRD events predicted a shorter time to bipolar depressive episode recurrence [35]. Individuals with BSD also were more susceptible than controls to SRD in response to equivalent life events [36, 37]. Although baseline social rhythm regularity and SRD events predict bipolar symptoms and episodes [33–35], whether this is due to circadian rhythm disruption has not been demonstrated. Repeated assessments of social, behavioral, and circadian rhythms in ambulatory participants over time, allowing for temporal precedence, stronger causal inferences, and examination of trajectories during adolescent development, are needed to test whether SRD and circadian rhythm disruption predict first onset of BSD and bipolar symptoms in adolescents’ daily lives. Project CREST is designed to provide such novel evidence.

In a separate literature, BSDs have been associated with an enhanced sensitivity to rewarding stimuli and higher reward-related brain function in cortico-striatal circuitry [9, 38, 39]. Moreover, high reward responsiveness (RR) is hypothesized to be a vulnerability for BSDs [9, 38–40]. According to these reward hypersensitivity models of BSDs, high trait RR can lead to excessive state increases in approach-related affect and behavior in response to life events that activate the reward system involving goal-striving or attainment, which in the extreme, is reflected in hypomanic symptoms or episodes (Fig. 1, red pathway). It also can lead to excessive state decreases in approach-related affect and behavior in response to events that deactivate the reward system involving definite failures and losses, reflected in bipolar depressive symptoms or episodes (Fig. 1, blue pathway). Reward hypersensitivity should make people hyper-reactive to cues signaling both the possible attainment and loss of reward, thus increasing risk for both hypo/mania and bipolar depression (vulnerability-stress hypothesis). From this perspective, mixed states may reflect rapid oscillations between increases and decreases in approach motivation in response to reward-activation and -deactivation cues, respectively, within the same day, leading to rapid switches between hypo/mania and depression that appear simultaneous in retrospective recall [9]. Alternatively, mixed states may involve the co-occurrence of reward-activation and –deactivation cues leading to a mixed motivational state and both hypo/manic and depressive symptoms simultaneously.

Much evidence supports the reward hypersensitivity model of BSDs (see [9, 38–40] for reviews). Individuals with or at risk for BSDs exhibit ambitious goal-striving [41–44], and greater behavioral, emotional, cognitive, and neurophysiological responses to rewards than controls [8, 45–49]. High self-report, behavioral, and/or neurophysiological RR predicts first onset and recurrences of BSD [8, 50], and progression to more severe bipolar diagnoses (e.g., bipolar I) among individuals with milder BSDs [7]. Further, reward-activation events involving goal-striving or attainment predict hypo/manic episodes among BSD participants [51–53], whereas reward-deactivation events predict depressive episodes [54–56]. Finally, fMRI studies provide partial support for elevated ventral striatum (VS) and orbitofrontal cortex (OFC) reward responses in manic and euthymic bipolar participants and participants at risk for BSD (see [39] for review, but see [57–60] for contrary results). Moreover, individuals at risk for BSD show greater cortico-striatal structural connectivity than low-risk participants [61] and those with a BSD exhibit greater gray matter volume in structures of the reward circuitry [62].

To test the reward hypersensitivity theory of vulnerability for BSD, a truly prospective, longitudinal study of first onset of BSD is required. Although a few studies examined baseline RR as a predictor of bipolar symptoms or episodes [7, 49, 50, 63], whether chronically high RR and increases in RR during adolescence predicts first onset of BSD is untested. This is important because the neural circuitry underlying reward function shows rapid maturation during adolescence [64, 65]. At least three timepoints are required to test whether developmental trajectories of RR predict first onset of BSD and increases in bipolar symptoms, as well as to test potential mediators of the RR – bipolar symptoms and BSD onset predictive associations. In addition, almost all research on the role of reward function in BSD has been limited to financial incentives. Given that social rewards increase in salience during adolescence [66–68], we also examine the trajectories of social reward processing as a predictor of first onset of BSD and increases in bipolar symptoms.

Furthermore, work on circadian and reward approaches to BSDs mostly has proceeded in parallel. Recently, however, integrated reward-circadian models of BSDs have been proposed [24, 69] based on evidence that the SCN and cortico-striatal reward circuitry are related bidirectionally [26, 69–73] and interact to affect mood processes [26, 55]. Indeed, RR exhibits time-of-day and circadian activation effects [70, 71, 74], clock genes are expressed in brain reward areas [75–77], and midbrain dopamine neurons project directly to the rodent SCN [78]. When regulated properly, reward-circadian interactions coordinate adaptive mood and behavioral functioning for goal pursuit. However, when either system becomes dysregulated, reward and circadian signaling may synergize to form a positive feedback loop, whereby dysregulation in one system exacerbates dysregulation in the other. Under these conditions, reward and circadian dysfunctions may interact to predict maladaptive mood and behavior, as reflected in hypo/manic and depressive symptoms (moderation hypothesis; arrows labeled “1” in Fig. 1). However, no prior study has tested whether RR interacts with circadian rhythm disruption to predict BSD onset and symptoms prospectively.

Project CREST provides the first systematic test of reward-circadian dysregulation as a joint vulnerability for increases in bipolar symptoms and first onset of BSD in the vulnerable period of adolescence. We use a biobehavioral high-risk design involving multilevel social and circadian rhythm measures (self-report, actigraphy, melatonin) and multilevel (self-report, behavioral, neural) and multimodal (monetary, social) RR measures in a prospective longitudinal design with repeated ecological momentary assessment (EMA) periods to examine: 1) concurrent and longitudinal bidirectional associations between social and circadian rhythm disruption and RR; 2) mediators of their associations, and 3) social and circadian rhythm dysregulation and RR as separate and joint predictors of increases in BSD symptoms and risk for first onset of BSD during adolescence.

Our hypotheses are as follows: 1) We predict that higher RR will be associated with greater social and circadian rhythm dysregulation (phase shift, lower amplitude, and less regular rhythms) concurrently and that increases in RR over time will predict greater social and circadian rhythm dysregulation over time and vice versa. 2) During EMA periods: a) reward-activation and -deactivation events will predict hypo/manic and depressive symptoms, respectively, mediated by SRD and circadian rhythm disruption; b) between-day changes in social and circadian rhythm disruption will predict hypo/manic and depressive symptoms, mediated by RR; and c) greater mood symptoms via these pathways will occur for participants selected for high trait RR. 3) Finally, we also predict that chronically high and increasing trajectories of RR and chronically high and worsening trajectories of social and circadian rhythm dysregulation will be separate and joint (interactive) predictors of increases in hypo/manic and depressive symptoms and first onset of BSD over time.

The goal is to recruit 320, 13–16 year-old adolescents to complete a prospective, 3-year longitudinal study. Recruitment began in Spring 2022. This age range was chosen based on several considerations. First, epidemiological studies of BSD incidence suggest that the first peak in BSD onset occurs between ages 14–19 [9, 82, 83]; thus, we wanted to recruit participants in the first half of this age range and follow them through most of this risk period. Second, RR normatively increases in adolescence and may start to peak around ages 15–16 [64, 65]. Third, sleep and circadian rhythms change normatively in adolescence [84–86]. Thus, this age range should provide the best opportunity to observe trajectories of RR-circadian rhythm associations and prediction of first onset of BSD.

Participants with no prior history of BSD are being selected along the entire dimension of self-reported trait RR, with oversampling at the high tail of the dimension to increase the likelihood of BSD onsets [8]. At Time 1 (T1) – Time 6 (T6), every 6 months, participants will complete self-report and interview assessments of life events coded for SRD and reward-relevance and self-report and psychiatric diagnostic interview measures of hypomanic and depressive symptoms and episodes for a total of 3 years of follow-up. Yearly, at T1, T3, and T5, participants also will complete self-report measures of circadian chronotype and social rhythm regularity, dim light melatonin onset (DLMO) saliva sampling procedures to assess circadian phase, self-report RR, goal-striving, and ruminative response styles, behavioral and neural RR (fMRI assessment of cortico-striatal [VS, OFC)] activation and functional connectivity during monetary and social reward tasks), and a 7-day EMA period. The behavioral and fMRI RR measures will occur on the day before the DLMO procedure and the start of the 7-day EMA period. During each 7-day EMA period, participants will complete continuous measures of sleep/wake and activity (actigraphy) and 3 within-day (morning, afternoon, evening) measures of life events coded for SRD and reward-relevance, monetary and social RR ratings, positive and negative affect, and hypo/manic and depressive symptoms. At T1 only, mothers also will provide information on their own psychiatric history, family psychiatric history, socioeconomic status (SES), and adolescent participants’ adversity history from birth to T1 (see Fig. 2). A STROBE Checklist [87] for this study may be found in Supplementary Materials.

A two-phase screening process is used to recruit adolescents from the community. First, they complete an online screening questionnaire including demographics questions, the Behavioral Inhibition System/Behavioral Activation System (BIS/BAS) [88] Scales, and contact information for themselves and a parent. The BAS total score from the BIS/BAS is a psychometrically reliable and valid measure of trait reward sensitivity [7, 8, 89] that we have used before to select participants with different levels of trait RR [8]. Based on their BAS scores, we plan to recruit 20 adolescents each from the 0-20^th, 21-40^th, 41-60^th, and 61-80^th quintiles, and 240 adolescents from the highest 80-100^th quintile of the RR dimension. The purpose of this recruitment design involving oversampling of participants in the highest quintile of BAS scores is to ensure that we have enough participants at elevated risk for first onset of BSD, while still maintaining representation of the full dimension of RR in the sample. To be eligible for the study, adolescents must be ages 13–16 and all racial and ethnic groups are eligible. To allow exploration of sex differences, we will try to recruit equal numbers of males and females.

Adolescents who are potentially eligible based on the online screener and their parents are then contacted to schedule a Phase II phone screening interview. The phone screening interview is used to describe the project to adolescents and their parents in detail and, if they are interested, to more fully determine whether the adolescent meets all eligibility criteria. The interview contains an MRI safety screening questionnaire, questions about eligibility for the circadian component of the study, and the current and past mood disorder sections of the Structured Clinical Interview for DSM-5 (SCID-5) [90]. The original source of this description of the phone screening interview may be found in Alloy et al. [91]. Eligible adolescents who wish to participate provide written assent and their parent provides written consent. Project CREST is approved by the Temple University IRB (#28338) and is being conducted in accordance with all relevant guidelines and regulations, including the Declaration of Helsinki.

We exclude adolescents with a history of a BSD (bipolar I, bipolar II, cyclothymia, or even a single hypomanic episode) because one of the aims of the project is to predict first onset of BSD in adolescence. To ensure validity of project assessments, we also exclude participants with psychotic symptoms (hallucinations, delusions). Standard MRI exclusion criteria (e.g., traumatic brain injury, pregnancy, metal in the body, severe claustrophobia) also are applied. In addition, adolescents are excluded based on the following: 1) use of melatonin or photosensitizing medications, 2) current or past treatment with bright light, and 3) shift work or travel across multiple time zones within the past month. These exclusion criteria are designed to ensure that there are no contraindications or major confounding factors for the DLMO saliva sampling procedure. If an adolescent has traveled across time zones, but is otherwise eligible, participation in T1 is delayed for at least 1 month. Psychotropic medication use is not a basis of exclusion; instead, analyses will control for medication use. This is designed to increase the sample’s representativeness and the generalizability of project findings, inasmuch as individuals with, and at risk for mood disorders often show elevated rates of psychotropic medication use [92].

Power analyses were conducted at the time of the NIMH grant application (MH126911) and prior to the start of data collection for Project CREST and informed sample size. All power estimates were derived using Monte Carlo simulations facilitated in Mplus 8.4 and were based on 1000 replications. For Hypothesis 1 cross-sectional associations, based on our planned sample size, we have strong power (> 0.90) to detect small (r = 0.20) bivariate associations. For longitudinal associations, we have good power (> 0.81) to detect small-moderate associations (r = 0.275) of prediction of slopes of one system from the intercept of the other within planned parallel process growth models.

For Hypothesis 2, within each EMA period, we will examine mediating relationships within the context of random-intercept cross-lagged panel (RICLP) models that include processes for each predictor, mediator, and outcome variable. Models focus on the indirect effects from the within-person assessments of the predictor to the outcome through the mediator variables and include complementary between-person associations to permit discriminating associations between levels of analysis. Within the context of these models, there is strong power (> 0.90) to detect small effects (r = 0.25) for between- and within-person associations. Further, power is strong (> 0.90) to detect significant indirect effects at both the within- and between-person levels of analysis. Moreover, we have good power (> 0.85) to detect small differences in the magnitude of indirect effects (i.e., standardized indirect effects differing by 0.06) in the context of the hypothesized interaction effects.

For Hypothesis 3, we have good power to detect small associations between growth factors of RR and social and circadian rhythm trajectories with symptom trajectories (> 0.85) and moderate power to detect moderate effects for prediction of BSD onsets (> 0.81). Based on findings from Alloy et al. [8], we estimate that 20.9% (N = 50) of the 240 participants from the highest quintile (80-100^th %) of the RR dimension, and 4.2% (N = 3) of the 80 participants from the rest of the RR dimension (0-79^th %), will develop a first onset of BSD by T6, for a total of 53 BSD onset cases. We also have adequate power (0.78) to detect moderate interaction effects (i.e., standardized regression coefficients differing by 0.30) predicting symptoms and similar power (0.82) for finding significant interaction effects for prediction of BSD onsets (i.e., odds-ratios for the interaction effect of approximately 1.6).

Study assessments and their timing are summarized in Tables 1 and 2. Baseline T1 and yearly T3 and T5 involve two sessions designed to occur within a week of each other (Session 1 is remote over videoconferencing; Session 2 is in person) and T2, T4, and T6, occurring 6 months after T1, T3, and T5 involve one remote session via videoconferencing (see Fig. 2).

The self-report, behavioral, and neural reward measures are given yearly at T1, T3, and T5.

Participants complete a series of self-report and objective measures of social/circadian rhythms and sleep yearly at T1, T3, and T5. To assess social rhythm regularity, participants complete the Social Rhythm Metric-Trait (SRM-T) [109]. This self-report measure seeks to quantify patterns of daily social behavior by assessing the regularity of the timing of 15 daily activities (e.g., bedtime, mealtime, exercise, etc.). Activities completed within 45 min of their average time are considered “regular” and a higher overall score reflects greater social rhythm regularity [109].

Chronotype, a trait-based indicator of circadian preference, is assessed using the Morningness-Eveningness Questionnaire (MEQ) [110]. This self-report measure assesses the degree to which individuals prefer morning or evening activities. Higher scores indicate greater morningness. Based on responses, individuals can be categorized as morning types (“morning larks”), evening types (“night owls”) or neither. The MEQ demonstrates good reliability [111] and is correlated with physiological markers such as dim light melatonin onset time [112, 113].

To assess sleep quality, participants complete the Pittsburgh Sleep Quality Index (PSQI), a 19-item self-report instrument that asks about the timing of sleep as well as sleep-related problems (e.g., insomnia, nightmares, breathing difficulties) over the past month [114]. This self-report measure consists of seven dimensions: subjective sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbances, use of sleep medication, and daytime dysfunction, as well as a total score that reflects overall sleep disturbance [114]. Psychometric investigations indicate the PSQI demonstrates strong reliability and validity across multiple samples, including a community sample of adolescents [115, 116]. The PSQI has been shown to discriminate between individuals with and without disorders associated with sleep disturbance [117].

The T1, T3, and T5 study assessments are split into two sessions, one week apart, and during the one-week interval participants complete one week of actigraphy, daily sleep diaries, and a dim-light melatonin onset (DLMO) procedure. Actigraphy is an objective, reliable, and valid method for assessing activity patterns, sleep/wake metrics, and light levels, with minimal restriction or interruption of normal routines or daily activities [118–121]. Actigraphy has been well-validated against polysomnography and sleep diaries among healthy sleepers and clinical samples [119, 122, 123].

To collect actigraphy data, participants continuously wear a Philips Actiwatch with light sensors for 7 days on their non-dominant hand, only removing it when it might get wet (e.g., bathing). Participants are instructed to keep their watch uncovered (e.g., roll their shirt sleeve) to maximize the accuracy of environmental light measurement. Data will be sampled in 1-min epochs and cleaned by trained research assistants using a standardized protocol in which study staff will visually scan the actograms to evaluate whether the rest intervals align with the light data and subjective sleep and make adjustments when there are clear inconsistencies. Sleep parameters will be extracted (total sleep time, sleep duration, wake after sleep onset [WASO], and midpoint of sleep using Philips’ proprietary algorithms. Four circadian rhythm rest-activity measures also will be derived from the actigraphy data: 1) acrophase (the time of peak activity level each day), 2) amplitude (the difference between the activity level acrophase and nadir each day), 3) sleep midpoint, and 4) social jetlag (the discrepancy between weekday and weekend sleep) [28, 124]. Social jetlag will be computed by measuring the difference in sleep midpoint on “free” days (e.g., weekend) and “work” days (e.g., school days; [125]).

Participants also fill out a sleep diary each evening to collect self-report information about the past night’s sleep and factors that may have impacted their sleep. Sleep diaries are considered the “gold standard” for subjectively assessing sleep [126] and sleep experts recommend using a sleep diary in concert with actigraphy to reduce uncertainty and collect additional relevant information [120]. The sleep diary includes questions assessing the time participants woke up that morning and the following characteristics of last night’s sleep: the time they got in bed, the time they tried to fall asleep, the time they actually fell asleep, and the number of times they woke up during the night. Subjective sleep disturbances are assessed by asking whether participants had trouble 1) falling asleep, 2) staying asleep, or 3) waking up too early and being unable to return to sleep. Participants report previous day medication usage, caffeine consumption (coffee, soda, energy drink, tea), and whether they napped (and if so, at what time and for how long). The sleep diary also includes a single item assessing the quality of sleep, “How well did you sleep last night?” and a question asking how sleepy or alert participants felt during the day. Participants respond to these items on a 10-point Likert scale.

Dim light melatonin onset (DLMO), the gold standard physiological marker of endogenous circadian function [127], is assessed on the first full day of the EMA week (i.e., the evening after the fMRI scan during the second session of T1, T3, and T5). The day of the DLMO procedure, participants must refrain from alcohol, bananas, chocolate, NSAIDs, beverages with artificial colors, and any caffeinated beverage after breakfast to avoid sample contamination. Thirty minutes before the start of sample collection, participants dim their indoor lights, use a smartphone application to ensure the ambient light level is less than 100 lx, and don light-attenuating glasses. Participants provide 11 saliva samples starting 6 h before their bedtime. Sampling frequency is one sample an hour for 3 h and then one sample every 30 min for 3.5 h such that the last sample is taken right before going to sleep. Participants eat dinner during the first 1-h intervals and are instructed not to consume food or water within 30 and 20 min of a sample, respectively. Passive drool saliva samples (at least 1 mL) are collected in cryovials using a saliva collection aid. The exact timing of the sampling is recorded in two places: on a physical sheet of paper and in the Salimetrics “OnTimePoint” smartphone application. Samples are stored in participants’ freezer until they are returned to the lab, where they are stored at -20 °C pending ELISA assay. DLMO is defined as the first interpolated point (derived from between 2 points) at 4.0 pg/ml on the rising curve of melatonin concentration and is an objective indicator of circadian phase [112, 128]. The minimum detectable limit of the assay is 0.5 pg/ml with intra- and inter-assay coefficients of variation of 6.8% and 7.3%, respectively.

Participants complete a modified version of the Life Events Scale (LES) [56] at T1-T6 to assess positive and negative, major and minor life events during the past 6 months. The original 193-item LES was shortened for Project CREST so that it contained 156 events in multiple domains (e.g., school, peers, romantic interests, family, financial) relevant to adolescents, including events particularly relevant to today (e.g., “Received a lot more positive attention in-person or on social media (e.g., more “likes” than usual),” “Tested positive for COVID-19”). Events were coded a priori as reward-relevant or not and into specific reward-relevant categories with inter-rater reliabilities of α’s = 0.79-0.94: Goal-Striving, Goal Attainment/Reward, Goal Obstacle, and Goal Failure/Loss [129]. The original source of the description of the LES may be found in Alloy et al. [91].

Following completion of the LES, adolescents complete a Life Events Interview (LEI) [56, 130] to obtain further information about the events endorsed on the LES and date when they occurred. Trained post-baccalaureate staff and clinical psychology doctoral students, blind to all study measures, conduct the interviews with adolescents. The LEI uses manualized, event-specific definitional criteria and probes to maintain consistency, avoid double-counting of events, and reduce dating errors. Events not meeting definitional criteria are disqualified to reduce subjective reporting biases. The LEI also employs the “gold-standard” contextual threat method [130] to rate the events’ objective impact on a scale from 1 (Mild) to 4 (Extreme) and independence (e.g., death of a family member) vs. dependence (e.g., fight with a friend) on the participant’s behavior. Social rhythm disruption (SRD) incurred from each qualifying event also is rated based on the number of hours the participant’s normal bedtime or wake time is changed by the event’s occurrence [35, 36, 55]. These procedures have yielded excellent reliability and validity of event dating and ratings (κ = 0.76 -0.89) in our previous studies [35, 36, 55, 56, 129]. The original source of the description of the LEI may be found in Alloy et al. [91].

The Children’s Life Events Scale (CLES) [131] is given to the parent and adolescent at T1 to assess negative events that the adolescent may have experienced throughout childhood. The parent also completes the CLES because there may be events that occurred when the adolescent was too young to remember. The CLES consists of a 50-item checklist including moderate to major negative events across the following domains: peer difficulties (e.g., “break up of serious romantic relationship”), achievement-related events (e.g., “academic failure”), family difficulties (e.g., “divorce of parents”, “serious financial difficulties of family”), maltreatment (i.e., emotional, sexual, physical), death of family member or close friend, and additional events suggesting inadequacy (e.g., “acquired a physical deformity”). For each event endorsed, the parent or adolescent reports the corresponding age of the adolescent at the time the event occurred. Scores for the CLES range from 0 to 50, with higher scores indicating a higher number of negative events, representing greater early adversity. Additional subset scores are derived from the total number of events reported within each subset category, including negative emotional feedback (i.e., “frequent teasing by peers”, “decrease in acceptance by peers”), family deaths (e.g., “death of a grandparent”, “death of a parent”), achievement failures (e.g., “academic failure”, “nonacademic failure”), events suggesting inadequacy (e.g., “acquired a physical deformity”, “needed special education services”), and dependent events and independent events. The CLES has shown predictive validity and good internal consistency (α = 0.75; [18, 131]).

At T1, both the teen and the parent complete the MacArthur Scale of Subjective Social Status [132, 133] in which they independently rank themselves on a 10-step ladder that represents their perception of their family’s social standing in the society. Lower rungs on the ladder indicate lower perceived social status whereas higher rungs indicate higher perceived social status. In addition, at T1, the parent is administered a childhood socioeconomic interview, which includes family structure, past and current income, past and current housing, parent education and employment status, and family immigration history. Family socioeconomic disadvantage is operationalized by taking the sum of the following indicators during both the first 5 years of the teen’s life and total lifetime (1 = present; 0 = absent): family poverty (i.e., income-to-poverty ratio < 1.00) [134], unemployment, single-parent family structure, parent non-completion of high school degree or equivalent, and immigration status. The original source of the description of the MacArthur Scale may be found in Alloy et al. [91].

At T1, T3, and T5, participants complete 7 days of EMA following an in-person session using the SEMA3 online platform [159]. During the 7-day EMA periods, in addition to completing a daily sleep diary and wearing an Actiwatch, participants complete affect, symptom, and RR ratings three times a day (morning, afternoon, evening). Both the EMA surveys and sleep diaries ask the participant to report any life events that have occurred since the last prompt. This EMA period begins the day after the in-person session, which includes the fMRI, and includes the DLMO procedure on the first evening of the EMA period. The season and day of the week that the EMA period begins are recorded as potential covariates.

At the first in-person session, participants download the SEMA3 smartphone application and receive training to understand how to respond to prompts. Participants also choose the most suitable time to receive the nightly sleep diary at their in-person visit before beginning the EMA periods (choices are 7, 8, 9, or 10 PM). Sleep diary surveys must be completed within 12 h. Generally, the EMA prompt window begins 12 h before the sleep diary survey time chosen by the participant. Participants are sent a morning, afternoon, and evening survey prompt within a 3-h window (e.g., 9AM-12PM, 1-4PM, and 5-8PM clock time), and each survey expires after an hour. In some cases in which participants’ schools strictly do not allow phone use, an alternative schedule is used. In these cases, the typical schedule is used for the weekend days in the 7-day period. For the school days, participants are sent the morning survey between 6 and 7:30 AM, the afternoon survey between 3 and 5 PM, and the evening survey between 6 PM and their chosen sleep diary time. These surveys also expire after an hour.

Adolescents who are fully eligible and wish to participate in the longitudinal project provide written informed assent and their parent or legal guardian provides written informed consent. Ethics approval was granted by the Temple University Institutional Review Board (IRB #28338) for this study. This study is being conducted in accordance with all relevant guidelines and regulations, including the Declaration of Helsinki.

Not applicable.

The authors declare no competing interests.

Data from this study will be available on the NIMH National Data Archive (Collection #C4089).