Prospective association between evening circadian preference and academic functioning in adolescents: the role of daytime sleepiness

Background There is growing evidence for the role of circadian factors in adolescents' sleep and academic adjustment, with greater evening preference being linked to poorer academic functioning. However, studies have yet to evaluate this association prospectively in adolescence, nor have studies examined daytime sleepiness as a putative mechanism linking evening preference to poor academic functioning. The current study used a multi‐informant design to test the prospective association of evening circadian preference, daytime sleepiness, and academic functioning (e.g., global academic impairment and grades) across 2 years in adolescence. As evening circadian preference, sleepiness, and academic problems are elevated in adolescents with ADHD, we used a sample enriched for adolescents with ADHD and explored whether ADHD moderated effects. Method Participants were 302 adolescents (M age = 13.17 years; 44.7% female; 81.8% White; 52% with ADHD). In the fall of eighth grade, adolescents reported on their circadian preference, and in the fall of ninth grade, adolescents and parents completed ratings of daytime sleepiness. In the middle of 10th grade, parents and teachers reported on adolescents' academic impairment and at the end of 10th grade, adolescents' grade point average (GPA) was obtained from school records. Results Above and beyond covariates (e.g., adolescent sex, ADHD status, medication, sleep duration) and baseline academic impairment, greater self‐reported evening preference in 8th grade predicted increased parent ratings of academic impairment in 10th grade indirectly via adolescent and parent ratings of daytime sleepiness in 9th grade. Furthermore, evening preference in 8th grade predicted greater teacher ratings of academic impairment and lower average GPA in 10th grade via parent ratings of daytime sleepiness in 9th grade, controlling for covariates and baseline GPA. ADHD status did not moderate indirect effects. Conclusion Findings underscore daytime sleepiness as a possible intervening mechanism linking evening preference to poor academic functioning across adolescence. Intervention studies are needed to evaluate whether targeting circadian preference and sleepiness improves academic functioning in adolescents.


Introduction
There has been growing interest in the role of circadian factors in the development and maintenance of sleep and academic functioning in adolescence (Arns, Kooij, & Coogan, 2021), a developmental period characterized by substantial biological, psychological, and social processes impacting sleep-wake regulation (Becker, Langberg, & Byars, 2015;Crowley, Wolfson, Tarokh, & Carskadon, 2018). Circadian preference, often assessed via self-report ratings, is characterized by the component of the sleep/wake cycle reflecting individuals' tendencies for the timing of waking activities and sleep (Jenni, Achermann, & Carskadon, 2005;Roeser, Schlarb, & K€ ubler, 2013). Circadian preference is considered a proxy for chronotype, which is assessed via biological indices (dim light melatonin onset), and research demonstrates significant associations between selfreported ratings of circadian preference and biological indices of circadian function (Zavada, Gordijn, Beersma, Daan, & Roenneberg, 2005). Additionally, studies report moderate individual stability of self-reported circadian preferences across adolescence (Kuula et al., 2018). Although adolescents undergo a natural biological shift toward later eveningness coinciding with pubertal development (Dolsen, Wyatt, & Harvey, 2019), adolescents vary on the morning-evening continuum. Specifically, adolescents with inclinations toward morning preference ('morning lark') generally prefer earlier sleep onset and waking activities, whereas those with an evening preference ('night owl') prefer later activities and sleep onset (Roeser et al., 2013). Across studies examining correlates of circadian preference, greater evening preference is linked to a host of emotional, social, sleep, and academic difficulties (Chen et al., 2021;Dolsen et al., 2019). Of particular concern, due to the lack of synchrony between later evening preference and expectations for early waking and school-based activities, is the detrimental impact on academic functioning (Tonetti, Natale, & Randler, 2015). Studies have linked later evening preference to lower average school grades (Cohen-Zion & Shiloh, 2018;Saxvig, Pallesen, Wilhelmsen-Langeland, Molde, & Bjorvatn, 2012) and increased academic difficulties (Tonetti et al., 2015). Recently, a recent meta-analysis of 23 studies reported a small yet significant association between greater evening preference and lower academic achievement (Scherrer & Preckel, 2021). However, the authors noted the absence of any longitudinal studies as a key limitation of extant research in this area.

Need for longitudinal studies
Longitudinal studies spanning adolescence are critical for informing our understanding of whether evening circadian preference is directly related to academic outcomes, as well as whether other developmental processes that may be important for understanding this association. To our knowledge, only one study to date has explored the prospective association of circadian preference and academic impairments (Scherrer & Preckel, 2021). In a community-based sample of adolescents, morning preference, but not evening preference, was associated with higher grade point average (GPA; Scherrer & Preckel, 2021) over a 2-year period. However, the two time point design of the study precluded the opportunity to examine intervenining processes that may explain the link between evening circadian preference and academic outcomes.
One particularly compelling candidate mechanism is excessive daytime sleepiness (Sivertsen, Glozier, Harvey, & Hysing, 2015). Adolescents with later evening preference report shortened sleep duration, poorer sleep quality, and increased difficulties waking (Giannotti, Cortesi, Sebastiani, & Ottaviano, 2002;Roeser et al., 2013;Vollmer et al., 2017). Additionally, greater evening preference has been linked to delaying bedtime and sleep onset (Chung, An, & Suh, 2020). In turn, adolescents often experience increased drowsiness and poor sustained attention in the early hours of the school day (Roeser et al., 2013). Furthermore, evening preference may have a stronger prospective association with the development of daytime sleepiness compared to other sleep domains such as quality and duration (Luo, Zhang, Chen, Lu, & Pan, 2018). Regarding the association between daytime sleepiness and academic problems, studies have linked greater daytime sleepiness to poorer academic performance (Chung & Cheung, 2008) and lower school grades (Liu et al., 2016;Meijer, 2008). In fact, systematic reviews and meta-analyses have observed stronger effects of daytime sleepiness on school performance compared to other types of sleep problems (Dewald, Meijer, Oort, Kerkhof, & B€ ogels, 2010;Tonetti et al., 2015). Thus, daytime sleepiness might be a key mechanism that explains the association of evening preference and academic outcomes in adolescence. Indeed, one cross-sectional study found daytime sleepiness to mediate the link between evening preference and global GPA in adolescents (Sivertsen et al., 2015).
In addition, although daytime sleepiness is directly linked to poor academic performance (Dewald et al., 2010;Drake et al., 2003;Meijer, 2008), very few studies have tested individual factors associated in the development of daytime sleepiness (Luo et al., 2018) or evaluated specific academic domains. For instance, studies assessing evening preference and daytime sleepiness with academic outcomes often measure global ratings of academic impairment or collapse across subjects when assessing grades (Dewald et al., 2010;Scherrer & Preckel, 2021;Sivertsen et al., 2015). Thus, to improve our conceptualization of how evening preference may be associated with academic outcomes, this study examines parent and teacher ratings of academic impairment in addition to overall GPA obtained from school report cards. We also examined the effect of evening preference and daytime sleepiness on specific academic grades in supplemental analyses.

Current study
This study builds on previous research by testing daytime sleepiness as an intervening mechanism linking evening circadian preference to academic impairments over 2 years in adolescence. Specifically, we tested whether self-reported evening preference in 8th grade was related to academic outcomes in 10th grade indirectly via 9th grade daytime sleepiness. In addition to global ratings of parent-and teacher-reported academic impairment, we collected school grades from report cards to provide a more specific and ecologically valid indicator of academic performance in adolescents (Scherrer & Preckel, 2021). As supplemental analyses given the absence of previous studies in this area, we also examined effects for specific academic subjects (i.e., english/language arts, history, math, science). Additionally, a multi-informant approach using both self and parent ratings of daytime sleepiness was used. The sample included adolescents with and without ADHD given the documented rates of later evening preference, daytime sleepiness, and academic problems in adolescents with ADHD (Becker, Kapadia, Fershtman, & Sciberras, 2020;Langberg et al., 2016;Lunsford-Avery, Krystal, & Kollins, 2016). This allowed us to evaluate whether the impact of circadian preference on daytime sleepiness and academic outcomes differed in an already at-risk group. Furthermore, we controlled for a number of variables that have been associated with sleep problems and academic impairments, such as medication use (Lunsford-Avery et al., 2016), sex (Arns et al., 2021), family income (Sivertsen, Bøe, Skogen, Petrie, & Hysing, 2017), and sleep duration (Becker, Epstein, et al., 2019). To our knowledge, only one study has tested the prospective association between evening circadian preference and academic impairments and found no direct effects of evening preference (Scherrer & Preckel, 2021). Thus, we did not make specific hypothesis for the direct effect. Consistent with recent recommendations in the literature (Rucker, Preacher, Tormala, & Petty, 2011), our second objective was to test the indirect effect regardless of results for the direct effect of evening preference on academic outcomes. Specifically, researchers have raised concern with Baron and Kenny's (1986) criterion for establishing a direct effect before testing an indirect effect primarily related to the inadvertently omitting intervening effects (Mathieu & Taylor, 2006;Rucker et al., 2011). Given literature showing greater evening preference to be associated with increased ratings of daytime sleepiness (Luo et al., 2018;Roeser et al., 2013) and sleepiness to poor academic outcomes (Liu et al., 2016;Meijer, 2008), we anticipated an indirect effect of evening preference on academic outcomes via increased daytime sleepiness. Finally, we explored ADHD group status as a moderator of these indirect effects.

Procedures
This study was approved by the institutional review board (IRB) at Cincinnati Children's Hospital Medical Center and Virginia Commonwealth University. This study uses data from three time points: the fall of 8th grade (T1), the fall of 9th grade (T2), and the fall/winter of 10th grade (T3). School staff at middle schools around Cincinnati, Ohio and Richmond, Virginia were provided with recruitment materials to distribute to families (letters/flyers to students, newsletters, e-mail blasts). Parents contacted the research staff in response to recruitment materials and, after meeting general phone screening criteria (e.g., enrollment in eight grade), were invited for the initial visit. At this visit, the adolescent and parent received a comprehensive assessment, with inclusion criteria including: full scale IQ >80 on Wechsler Abbreviated Scale of Intelligence, Second Edition (WASI-II) and meeting criteria for ADHD or comparison group. Adolescents eligible for the ADHD group met fifth edition of the Diagnostic and Statistical Manual for Mental Disorders (DSM-5) criteria for ADHD combined presentation or inattentive presentation based on the parent version of the Children's Interview for Psychiatric Syndromes (P-ChIPS; Weller, Weller, Teare, & Fristad, 1999). Adolescents were eligible for the comparison group if the parent endorsed <4 ADHD symptoms for both inattentive and hyperactiveimpulsive domains. Exclusionary criteria included adolescents being previously diagnosed with or meeting criteria for autism spectrum disorder, bipolar disorder, dissociative disorder, a psychotic disorder, or an organic sleep disorder, or families who do not have sufficient English language to complete study materials. These exclusionary criteria resulted in 11 families being ineligible for the study. Parents provided written consent and adolescents provided assent at the first visit.
Diagnostic assessment. Adolescents were evaluated for ADHD at the initial visit in accordance with the DSM-5 criteria. The parent version of the P-ChIPS (Weller et al., 1999) was used to determine adolescent eligibility for the ADHD group. Adolescents were required to meet all DSM-5 criteria for either ADHD combined presentation or predominately inattentive presentation according to parent report on the P-ChIPS. Adolescents participated in the comparison group if parents endorsed ≤4 symptoms of both inattention and hyperactivity/ impulsivity.

Measures
Sleep habits survey (SHS). The SHS (Wolfson & Carskadon, 1998) is a self-report measure of circadian preference and sleep-wake difficulties. The 10-item circadian preference subscale assesses preference for morningness or eveningness ('When does your body start to tell you it is time for bed'?), with lower scores indicating greater evening preference. The circadian preference subscale has established internal consistency and convergent validity with ratings of later sleep onset and waketime (Giannotti et al., 2002;Russo, Bruni, Lucidi, Ferri, & Violani, 2007). Additionally, adolescents reported on their average sleep duration during the schoolweek. Internal consistency for circadian preference was a = .77.
Pediatric daytime sleepiness scale (PDSS). The PDSS was adapted from the adult Epworth Sleepiness Scale for assessing adolescents' daytime sleepiness (Drake et al., 2003). Adolescents reported on their likelihood of falling asleep in different situations (e.g., while doing homework), extent of sleepiness or drowsiness ('How often do you get sleepy or drowsy while doing homework?'), and difficulties waking on a 5-point scale (0 = never, 4 = very often). Parents also provided their perception of adolescents' daytime sleepiness ('How often does your child get sleepy or drowsy while doing homework?'). The PDSS has well-documented psychometric properties and convergent validity with sleep problems (Drake et al., 2003;Liu et al., 2019). Internal consistencies for self and parent-reported daytime sleepiness were as = .84 and .82, respectively. Impairment rating scale (IRS). The IRS is a widely used rating scale of perceived impact of behaviors on domains of impairment (Fabiano et al., 2006). The IRS has demonstrated adequate test-retest reliability and convergent validity with measures of impairment across community and clinical samples (Fabiano et al., 2006). Parents ('How your child's problems affect his or her academic progress at school') and teacher ('How this child's problems affect his or her academic progress') are instructed to endorse on a 7-point scale the extent of adolescents' problems in academics, with higher scores reflecting greater academic impairment and need for treatment in this area. Circadian preference obtained at the end of 8th grade and the end of 10th grade. Grades were quantified on a 4.0 grading scale consisting of a composite score of these subjects. Specific subject grades were also used in supplemental analyses.
Covariates. Parents were administered an adaptation of the Services Use in Children and Adolescents-Parent Interview (SCA-PI; Hoagwood et al., 2004) to assess medication use (for ADHD, sleep [including melatonin], and/or an emotional/ behavioral problem [e.g., antidepressants]). Medication use was coded as follows: 0 = no or unsure of medication status, 1 = yes. Parents also reported on family income, child's sex, and child's age.

Analytic strategy
Significant demographic factors associated with T3 academic impairment ratings and GPA were used as covariates. Path analyses and tests of indirect effects were conducted in the statistical software Mplus (version 8.1). Manifest variables were used given the sample size, and all models were just identified (df = 0). Consistent with recommendations on handling nonignorable missing data, we used full-information maximum likelihood with adolescent age and family income as auxiliary variables as an approach to reduce the impact of the nonignorable missing data biased parameter estimates (Graham, 2009;Nicholson, Deboeck, & Howard, 2017). We tested three indirect effect models, with T1 self-reported circadian preference as the predictor, T2 self-or parent-reported daytime sleepiness as the mediator, and T3 parent-reported academic impairment, teacher-reported academic impairment, and GPA as outcome variables in the separate models. We controlled for adolescent sex, age, ADHD group status, T1 medication status, T1 selfreported sleep duration, and baseline academic impairment or GPA (in each model, respectively). The direct and indirect effects and their 95% confidence intervals (CIs) were calculated using 10,000 bias-corrected bootstrapped sampling estimates (Mallinckrodt, Abraham, Wei, & Russell, 2006). Furthermore, aligning with recent recommendations on mediation (Rucker et al., 2011), we tested the indirect effect of these models even in the absence of a total or direct effect of T1 circadian preference on T3 academic impairment. Finally, as an exploratory analysis, we tested whether the indirect effect was moderated by ADHD group status (0 = comparison, 1 = ADHD).

Participant attrition
There was a small-to-moderate amount of participant attrition across the three time points (7% for T2 daytime sleepiness measures; 12% for T3 parentreported academic impairment; 30% for teacherreported academic impairment). Furthermore, 3.6% of participants did not have GPA data at baseline, whereas 22.8% did not have GPA at the T3 visit.
Little's MCAR test in SPSS indicated that data were missing in a pattern consistent with not missing at random (v 2 (125) = 196.276, p < .001), with participants completing all time points being older, having a higher parent-reported family income, and more likely to be in the non-ADHD group (ps < .05). Thus, adolescent age and family income were used as auxiliary variables in missing data estimation moving forward. Primary study variables were normally distributed (skew <2.00, kurtosis <4.00). Table 1 presents zero-order correlations and descriptive statistics between demographic variables and primary study variables. The following demographic factors were significantly correlated with T3 ratings of academic impairment and GPA: ADHD group status, older adolescents (only for academic impairment), adolescent girls (only for academic impairment), and families reporting lower household income (ps < .05). These variables were included as covariates in primary analyses, with age and family income as an auxiliary variable in missing data estimations.

Bivariate associations
Analyses testing the indirect effect of evening circadian preference on academic impairment via daytime sleepiness The first set of analyses evaluated relations between T1 evening preference, T2 self or parent ratings of daytime sleepiness, and T3 parent-reported academic impairment. Regarding effects of specific paths, T1 evening preference was uniquely associated with increased T3 parent-reported academic impairment (b = À.117, p = .014), controlling for covariates and T1 parent-reported academic impairment. ADHD status was associated with higher T3 ratings of academic impairment (b = À.184, p = .014). Additionally, T1 evening preference was uniquely associated with greater T2 self (b = À.335, p < .001) and parent (b = À.307, p < .0001) ratings of daytime sleepiness. Adolescent girls had greater T2 self-reported daytime sleepiness (b = .122, p = .031), and T1 medication status (b = .152, p = .034), shorter sleep duration (b = À.142, p = .016), and parent-reported academic impairment (b = .206, p = .004) were each significantly correlated with greater T2 parent-reported daytime sleepiness. Finally, T2 self-(b = .101, p = .052) and parent-reported daytime sleepiness (b = .147, p = .008) were uniquely associated with greater parent-reported T3 academic impairment. Second, we tested T1 evening preference, T2 self or parent ratings of daytime sleepiness, and T3 teacher-reported academic impairment. Consistent with bivariate correlations, T1 evening preference was unrelated to T3 teacher-reported academic impairments. Similar to the previous model with parent ratings of academic impairment, T1 evening preference was uniquely associated with both T2 self (b = À.338, p < .001) and parent (b = À.331, p < .001) ratings of daytime sleepiness. Similarly, T1 medication status (b = .152, p = .034) and shorter sleep duration (b = À.142, p = .016) were associated with greater T2 parent-reported daytime sleepiness. T2 parent-reported daytime sleepiness (b = .165, p = .030), but not self-reported sleepiness (p > .05), were uniquely associated with increased T3 teacher-reported academic impairment.    Finally, as presented in Table 2, we examined the indirect effect of T1 evening preference ? T2 self-or parent-reported daytime sleepiness ? T3 academic impairment (for parent and teacher ratings, separately). The indirect effect of T1 evening preference ? increased T3 parent-reported academic impairment via greater T2 self and parent ratings of daytime sleepiness was significant. Furthermore, the indirect effect of T1 evening preference ? increased T3 teacher-reported academic impairment via greater T2 parent-reported daytime sleepiness was significant. Conversely, the indirect effect via self-reported sleepiness was nonsignificant (p > .05).

Indirect effect of evening circadian preference on student grades via daytime sleepiness
Above and beyond covariates and baseline GPA (collected end of eighth grade), there was no direct effect of T1 evening preference on T3 GPA. Similar to the above models, T1 evening preference was uniquely associated with greater T2 self (b = À.333, p < .001) and parent (b = À.301, p < .001) ratings of daytime sleepiness. Furthermore, T2 parentreported daytime sleepiness (b = À.120, p = .020), but not self-reported ratings of sleepiness (p > .05), was uniquely associated with lower T3 GPA, controlling for baseline covariates.
As supplemental analyses, we examined the direct and indirect effect on specific academic subjects. First, controlling for covariates and baseline grades, findings showed a direct effect of T1 evening preference on lower T3 science grades (b = À.140, p = .017), but not T3 english/language arts, history, or math grades (ps > .05). Furthermore, controlling for baseline covariates and grades, T2 parentreported daytime sleepiness was associated with lower T3 History (b = À.128, p = .033) and Science (b = À.150, p = .011) grades, but not english or math grades. T2 self-reported daytime sleepiness was unrelated to T3 grades (ps > .05). Finally, as presented in Table 3, we then tested the indirect effect of T1 evening preference ? T2 self-or parent-reported daytime sleepiness ? T3 grades. The indirect effects of T1 evening preference on lower T3 english and history grades through T2 parent-reported daytime sleepiness were significant. None of the indirect effects via self-reported daytime sleepiness were significant.

Moderated indirect effect
Across all models, ADHD group status did not moderate the indirect effect of circadian preference on academic impairment, overall GPA, or specific academic subjects (ps > .05). 1

Discussion
The current study used a multi-method (e.g., rating scale and GPA in core academic subjects obtained from school report cards), multi-informant (e.g., self and parent ratings) design to test the prospective association of evening circadian preference, daytime sleepiness, and academic outcomes across 2 years in a sample of adolescents with and without ADHD. First, above and beyond demographic factors, ADHD group status, medication status, sleep duration, and baseline ratings of academic outcomes, factors shown in previous research to be associated with academic outcomes (Arns et al., 2021;Becker, Langberg, et al., 2019), adolescents' evening preference in the fall of 8th grade was uniquely associated with increased parent-reported academic impairment in the middle/end of 10th grade. To our knowledge, these findings build on cross-sectional studies (Saxvig et al., 2012;Sivertsen et al., 2015;Tonetti et al., 2015) by documenting prospective associations with ratings of academic impairment. With the exception of direct effects on lower science grades in 10th grade, evening preference was not directly related to teacher-reported impairment, overall GPA, or other academic subject grades. These findings are consistent with the only longitudinal study to date finding morning circadian preference, but not evening preference (assessed on separate dimensions), to be directly associated with All analyses also controlled for ADHD group status, sex, T1 medication status, T1 sleep duration, and baseline ratings of academic impairment. Standardized ab indirect effect and unstandardized confidence intervals are presented. increases in overall GPA 2 years later (Scherrer & Preckel, 2021). However, consistent with recent recommendations for testing intervening mechanisms in the absence of direct effects to better understand developmental processes (Rucker et al., 2011), we sought to evaluate daytime sleepiness as a process that explain the indirect effect of evening preference on academic outcomes.

Indirect effect via parent ratings of daytime sleepiness
In contrast to the limited direct effects, a consistent pattern emerged in which evening preference was indirectly associated with both parent and teacher ratings of academic impairments and lower overall GPA via parent ratings of daytime sleepiness. Importantly, these indirect effect findings were significant for adolescents' grades in 10th grade derived from school report cards, which is an important marker of academic achievement that strongly influences educational decision and success (Arbabi, Vollmer, D€ orfler, & Randler, 2015;Scherrer & Preckel, 2021). These findings highlight the importance of considering mechanisms in studies examining how evening preference longitudinally influences functioning (Scherrer & Preckel, 2021). Adolescents with an evening preference experience a mismatch between sleep timing, optimal performance, and early school start times (Dolsen et al., 2019). In turn, the accompanying difficulties with morning waking and sleepiness during early portions of the school day likely have a cascading effect on academic performance. Indeed, experimental evidence demonstrates that adolescents with evening preferences perform poorly on cognitive measures earlier in the day (Goldstein, Hahn, Hasher, Wiprzycka, & Zelazo, 2007). Interestingly, it was parent perceptions of adolescents' daytime sleepiness that consistently emerged as a mechanism of the link between evening preference and academic outcomes. Associations between parent ratings of adolescents' daytime sleepiness and academic impairment and grades were noticeably stronger compared to self-reported ratings, consistent with a meta-analyses finding larger effects of sleepiness on academic outcomes in studies using parent or objective measures of academic performance compared to self-report (Dewald et al., 2010). Additionally, parents may be equipped to observe the impact of sleepiness on academic difficulties across contexts, such as observing heightened drowsiness upon morning waking and during homework completion (Owens, Dearth-Wesley, Lewin, Gioia, & Whitaker, 2016).
Findings from supplemental analyses suggested that evening preference was directly associated with lower science grades and indirectly related to lower english and history grades via daytime sleepiness. Before interpreting these findings, we note that the significant direct effect of T1 evening preference on T3 math (per correlation results) and science (per correlation and direct effect results) grades were small in magnitude. Furthermore, findings from the direct and indirect effect of evening preference on specific grades were not consistent and findings should be interpreted cautiously. Still, a few studies have shown unique effects of sleep disturbances and sleepiness on verbal abilities and complex cognitive functions compared to nonverbal/automatic processes (Buckhalt, El-Sheikh, & Keller, 2007;Macchitella, Marinelli, Signore, Ciavolino, & Angelelli, 2020), whereas other studies find elevated sleepiness to undermine working memory and processing speed abilities (Calhoun et al., 2012) and impact math performance (Macchitella et al., 2020). Our findings for specific subject areas need to be replicated before drawing conclusions, and additional prospective studies evaluating circadian preference on specific academic subjects are needed to shed light on possible underlying neurocognitive or learning mechanisms.
Finally, the magnitude of indirect effects did not differ for adolescents with and without ADHD. Although surprising at first given documented elevations in evening preference, daytime sleepiness, and academic impairments in adolescents with ADHD (Langberg et al., 2016;Lunsford-Avery et al., 2016), these findings simply mean that evening preference and daytime sleepiness do not differentially or disproportionately impact the academic outcomes of adolescents with ADHD. Indeed, studies have found evening preference to be associated with daytime sleepiness and academic impairments across general community-based samples of adolescents (Luo et al., 2018;Roeser et al., 2013;Scherrer & Preckel, 2021). Collectively, our findings highlight the role of daytime sleepiness as an intervening mechanism linking circadian evening preference with academic outcomes. With few exceptions, these findings indicate that evening preference and academic outcomes may not be related directly but through their relationship with daytime sleepiness (Mathieu & Taylor, 2006). Moving forward, as research continues to investigate circadian function and daytime sleepiness on academic adjustment, it will be critical for longitudinal studies to explore specific underlying neurocognitive and academic processes underlying this developmental sequence. Regarding the synchrony of circadian preference and peak performance (Goldstein et al., 2007), increased daytime sleepiness may result in disruptions on attention regulation, learning or motivation, homework completion, or specific neurocognitive processes (e.g., verbal comprehension, processing speed; Buckhalt et al., 2007;Sivertsen et al., 2015). Future research, potentially using ecological momentary assessment or daily diary studies, would be well equipped to elucidate whether the impact is limited to the morning period or cascades into later periods and classes during the school day.

Limitations
Despite a number of strengths, including a multiinformant examination of the longitudinal association of circadian preference, daytime sleepiness, and academic impairment over 2 years in adolescents with and without ADHD, a few limitations are important to note. First, our prospective observational study precludes drawing causal influences of evening preference on sleepiness and academic performance. Thus, we encourage future research to consider implementing experimental designs to better understand mechanisms linking evening preference to increased sleepiness and, in turn, poor academic outcomes. Specifically, similar to studies examining effects on cognitive abilities (Goldstein et al., 2007), randomly assigning adolescents with late circadian preference to complete academic work at different times of the day or at times when circadian rhythm aligns with peak (or poor) performance. Second, our measure of evening circadian preference is not a direct measure of circadian function, and future research should include biological indices of circadian functioning such as dim light melatonin onset (Zavada et al., 2005). Importantly, objective measures of chronotype often find intraindividual variance across adolescence (Karan et al., 2021), pointing to the importance of repeated assessments and to better understand if evening preference stability and change is associated with sleepiness and academic outcomes. Along these lines, subjective ratings of daytime sleepiness introduce risk of bias, and future research is encouraged to consider objective or naturalistic designs to capture daytime sleepiness and drowsiness. Second, we did not have information related to school start times and whether this moderated indirect effects, which is a key area for future research. School timing has been found to influence the association between preference and academic performance (Goldin, Sigman, Braier, Golombek, & Leone, 2020), though contradictory findings have been reported as well (Martin, Gaudreault, Perron, & Laberge, 2016). Finally, participants who completed ratings across the three time points had higher family income and were more likely to be in the non-ADHD group; thus, future research is encouraged to include adolescents of diverse socioeconomic and racial backgrounds to further understand the effect of evening preference on sleepiness and academic outcomes.

Conclusion
Findings from the current study are the first to identify daytime sleepiness as a mechanism linking evening circadian preference and academic outcomes across a 2-year period in adolescence. Specifically, self-reported evening circadian preference was associated with increased parent ratings of academic impairment through self-and parent-reported daytime sleepiness, whereas parent ratings of daytime sleepiness explained the link between evening preference and increased teacher-reported impairment and lower GPA, in addition to lower grades in english/language arts and history subjects specifically. These findings highlight the continued need for longitudinal mediation studies to better understand how circadian factors impact academic adjustment in adolescents.

Key points
This study tested the prospective association between adolescents' circadian evening preference, daytime sleepiness, and academic outcomes.
A sample of adolescents with and without ADHD was included.
Self-reported evening preference in 8th grade predicted 10th grade parent-reported academic impairment via adolescent and parent ratings of daytime sleepiness in 9th grade.
Evening preference in 8th grade predicted greater teacher ratings of academic impairment and lower GPA scores in 10th grade via parent ratings of daytime sleepiness in 9th grade.
Indirect effects were significant controlling for baseline academic functioning and covariates.
Daytime sleepiness is a likely mechanism linking evening preference to academic outcomes in adolescence.
Notes 1. Given differences in type of T1 medication status (31% ADHD medication; 12% sleep medication; 9% medication for emotional/behavioral problems), we decided to re-run analyses with each medication type as a covariate separately. Results for the indirect effect and moderated indirect effect were unchanged.