Ours is the first study to determine and compare concurrent predictors (external correlates at the same point in time) of CDS controlling for multiple relevant variables during childhood (mean age 8.7 years) and during adolescence (mean age 16.4 years) and to compare these findings with longitudinal predictors of CDS over an 8-year period from childhood to adolescence in a population-based sample. Multiple independent variables were analyzed including parent ratings of psychopathology, sleep problems, and somatic complaints; psychological test scores (IQ, achievement, and neuropsychological); and demographics. This developmental analytic approach provides a greater understanding of CDS and possible external correlates and etiologic factors, which may help to identify preventive strategies and develop targeted interventions.
Concurrent Predictors
Significant concurrent predictors of CDS at both baseline and at follow-up 8 years later were remarkably consistent, demonstrating the similarity of CDS external correlates in childhood and in adolescence. Parent ratings of autism symptoms, insomnia symptoms, inattention, somatic complaints, and excessive sleep were positive independent predictors at both time points. Additionally, the baseline hyperactivity/impulsivity score was a significant negative predictor of the baseline CDS score and the follow-up depression score was a significant positive predictor of the follow-up CDS score. These external correlates of CDS in childhood and adolescence have been reported in other multivariate studies analyzing data at single points in time [7, 8, 11, 15, 17, 18]. Interestingly, the six significant follow-up concurrent predictors of CDS explained considerably more of the variance (50%) than did the six baseline concurrent predictors (29%). This is not explained by the addition of depression as a follow-up CDS predictor (which was the least significant of the predictors and only increased explained variance by 0.7%). It appears that external correlates of CDS at a mean age of 16 years are more strongly related to CDS than at a mean age of 9 years.
Longitudinal Predictors
Two of the significant concurrent predictors of CDS in our study (autism and insomnia symptoms) were significant longitudinal CDS predictors, along with baseline CDS scores. As expected (due to changes in symptoms over time), significant longitudinal predictors of CDS were fewer and explained less of the variance (18%) than concurrent predictors at baseline (29%) and follow-up (50%). Not surprisingly, the baseline CDS score was the most powerful predictor of CDS 8 years later, indicating that CDS during the elementary school years is the strongest risk factor for adolescent CDS compared with the other relevant variables examined in our study. However, baseline CDS explained only 11% of the variance in follow-up CDS scores. A previous study using the same sample and instrument as the current study showed considerable individual instability of CDS over the 8-year period. Of those with CDS at baseline, 63% no longer had CDS at follow-up, and 71% of those with CDS at follow-up were new cases and did not have CDS at baseline [12]. Interestingly, these remission and new case rates were similar to rates for other symptom scores in the same study, including inattention, impulsivity, oppositional behavior, irritability/temper outbursts, conduct problems, anxiety, and depression [12]. Much remains to be learned about longitudinal predictors of CDS and other psychological problems and why these problems are stable in some children and not in others.
Autism
Parent ratings of autism symptoms were the second most powerful longitudinal predictor of follow-up CDS scores (after baseline CDS ratings) and were also significant concurrent predictors of CDS at baseline and at follow-up. Results demonstrate that the presence of autism symptoms increases the risk of CDS concurrently and longitudinally. These findings are not surprising given the high prevalence of CDS in children with autism, with estimates ranging from 30%-49% [6, 29–31]. These estimates are slightly higher but similar to 27%-40% reported for ADHD [6, 17, 32–34]. Despite this, most CDS research is conducted with ADHD and not autism samples and most multivariate studies control for ADHD and not autism symptoms. Research shows that autism is a significant independent correlate of CDS that is not explained by its association with ADHD [6, 7]. The importance of our finding that autism (and not other psychopathology including ADHD) is a significant independent longitudinal predictor of CDS cannot be overstated and emphasizes the importance of considering and controlling for autism symptoms in future research. Clinicians need to be aware that autism, not only ADHD, increases the risk of CDS.
Sleep Problems
Parent ratings of baseline insomnia symptoms (sleeps less than most other children, trouble falling asleep, wakes often at night, and wakes too early in the morning) represented the third strongest and final significant longitudinal predictor of follow-up CDS and was also a significant concurrent predictor of CDS at baseline and at follow-up. In a recent commentary [35], Sadeghi-Bahmani and Brand stressed the need for longitudinal research controlling for relevant variables (including psychopathology related to both sleep problems and CDS) to determine a possible causal association between sleep problems and CDS. Our study is the first to conduct such analyses and showed that the baseline CDS score was not a significant predictor of follow-up insomnia symptoms, whereas baseline insomnia symptoms predicted follow-up CDS. This suggests that the relationship between CDS and insomnia symptoms is not bidirectional (i.e., childhood insomnia symptoms predict adolescent CDS, but childhood CDS does not predict adolescent insomnia symptoms when other relevant variables are controlled). Interestingly, only baseline hyperactive-impulsive and autism symptoms significantly predicted follow-up insomnia symptoms. The finding that baseline hyperactivity-impulsivity and autism symptoms are predictive of later insomnia is not surprising given that ADHD-C and autism are often chronic conditions and both are strongly associated with insomnia symptoms [36–39].
Previous studies controlling for other variables have demonstrated an independent positive relationship between parent report of sleep problems and CDS in children at the same point in time using total sleep problems scores [7, 33, 40–42]. Some studies have used multivariate analyses to determine specific types of sleep problems that are independently related to CDS. A novel study using adolescent self-report of CDS and objective and subjective sleep measures (actigraphy, daily sleep diary, adolescent and parent sleep ratings, and adolescent report of circadian preference) controlling for demographics, pubertal development, medication use, and ADHD status showed that CDS was uniquely associated with shorter sleep duration, later sleep onset, difficulty falling and staying asleep, difficulty awakening in the morning, daytime sleepiness, and later eveningness preference in adolescents with and without ADHD [43].
In other multivariate studies, parent report of reduced sleep (sleeps less than normal, wakes too early, and difficulty falling asleep) was a significant independent correlate of CDS in elementary school children, whereas fragmented sleep (talks or walks in sleep, wakes during the night, nightmares, and restless sleep) was nonsignificant [8]. A sleep disordered breathing symptom score was an independent correlate of CDS in a parent survey of 1,628 6- to 10-year-olds [44]. Among college students, self-report of subjective sleep quality and sleep disturbance (e.g., bad dreams, waking during the night, and feeling too hot or too cold) were significantly associated with CDS controlling for other psychopathology, whereas sleep latency, sleep duration, sleep efficiency, and use of sleep medication were nonsignificant [45]. Taken together, all of these results along with our finding of a link between CDS and insomnia symptoms, suggest that sleep problems in children and adolescents with CDS are multi-faceted and that study results are somewhat inconsistent, perhaps because of methodologic differences in assessment and differences in sample characteristics (e.g., age and diagnoses). However, across studies, insomnia symptoms appear to be most strongly linked with CDS.
In our study, excessive sleep was an independent concurrent predictor of CDS at baseline and at follow-up and was a stronger concurrent predictor than insomnia symptoms, which was also a significant predictor. Similarly, in 2,492 children with autism and children with ADHD, parent report of sleeping more than normal and a total sleep problems score (difficulty falling asleep, wakes during the night, restless sleep, talks or walks in sleep, nightmares, wakes too early, and sleeps less than normal) were significant independent concurrent predictors of CDS, with excessive sleep a stronger predictor than sleep problems [7]. This suggests that both types of sleep problems are related to CDS. CDS may reflect a disturbance in arousal for some individuals, specifically hyperarousal in the case of insomnia and hypoarousal in the case of excessive sleep (in that children may be physiologically sleepier throughout the 24-hour circadian cycle day and night). Importantly, however, from a longitudinal perspective, the only sleep-related independent predictor of CDS in our study was insomnia symptoms, suggesting that insomnia symptoms, unlike excessive sleep, may represent a premorbid risk factor for CDS rather than a manifestation of the syndrome. This may have clinical significance in that effectively treating insomnia in childhood may reduce the likelihood of CDS in adolescence.
Becker and Willcutt [46] speculate that a dysregulation in arousal may also be relevant to understanding CDS, given CDS’s association with affective symptomatology (hyperarousal) and daytime sleepiness and ADHD-Inattentive symptoms (underarousal). Although CDS and daytime sleepiness are significantly related [41, 43, 46], self- and parent-report studies show that daytime sleepiness is distinct from CDS in elementary school students [44], middle school students with ADHD [47], college students [48], and children referred to a sleep clinic [40]. Interestingly, daytime sleepiness was significantly related to sleeping more but not to sleeping less than normal in children with ADHD and community controls [38] and in children with autism [49], suggesting that children who are sleepier during the day are also sleepier at night.
Another more recently reported pathway to explore is the relationship between CDS and the tendency toward an evening circadian preference (most alert and productive during the evening or night hours or an individual characterized as a night owl). Evening circadian preference may share clinical facets with CDS, as it does with insomnia, particularly in individuals with a misalignment between their internal body clock and external environment that affects timing and duration of sleep. A study of adults showed that those with an evening preference had greater CDS symptoms than those with a morning preference [50]. Fredrick et al. [43] reported an independent association between CDS and later evening circadian preference, raising the possibility that CDS might represent an underlying circadian phase delay and that circadian preference may contribute to CDS. Fredrick et al. hypothesized that eveningness, circadian rhythm, and underarousal might be potential biomarkers of SCT. Lunsford-Avery et al. [51] also found an association between CDS and evening diurnal preference in adult psychiatric outpatients seeking an ADHD evaluation. Lunsford-Avery et al. concluded that it is not clear if eveningness represents a cause, consequence, or correlate of CDS and speculated that eveningness could contribute to sleepy and sluggish daytime behavior, that CDS symptoms and difficulty completing work during the day could delay bedtime, and that CDS and eveningness could co-occur, possibly due to a shared underlying neurobiological or genetic etiology. The authors further noted that eveningness is a close correlate of delayed circadian rhythm, which may be a biomarker of CDS [51].
In addition to the need for longitudinal research, the recent commentary by Sadeghi-Bahmani and Brand [35] called for EEG studies to better understand a possible causal relationship between insomnia and CDS. Notably, however, objective overnight polysomnography (PSG) scores assessing physiologic sleep and its stages were not associated with CDS in 665 elementary school children [8]. Correlations were all small (-.09 to .08, explaining < 1% of the variance) between CDS and the 14 PSG scores, including EEG arousals, sleep onset latency, awakenings, sleep efficiency, proportion of REM and nonREM sleep stages, REM latency, frequency and severity of snoring, number of apnea or hypopnea events, and mean oxygen saturation percentage. An important potential future direction is using power spectral EEG analysis to better understand the microstructure of the sleep EEG, its trajectories across the transition from childhood to adolescence, and its relationship to CDS.
The findings of significant relationships between CDS and parent-reported sleep problems (but not between CDS and PSG measures) could be explained to some degree by shared variance (i.e., same informant and instrument for subjective but not objective data). The discrepancy in findings based on PSG versus parent-reported sleep data in children has been reported in other research [25, 52, 53], and correlations between adult self-reported sleep problems and PSG data are poor [54]. Future research utilizing objective and subjective measures, as in studies by Fredrick et al. [43] and Mayes et al. [8], is important to investigate CDS and the arousal/circadian domains. Specifically, studies using multiple sleep latency tests to assess physiologic sleepiness are needed. Further, more studies using actigraphy to assess habitual sleep-wake patterns are necessary to build upon the finding by Fredrick et al. [43] that CDS was significantly correlated with actigraphy measures of later sleep onset and shorter sleep duration but not sleep efficiency. Much remains to be learned about specific aspects of sleep problems that are associated with CDS and why they are related.
Inattention
Although baseline inattention was not a significant independent predictor of follow-up CDS scores, inattention was a significant concurrent predictor of CDS at baseline and at follow-up. ADHD-Inattentive symptoms are most often cited as a correlate of CDS in the literature, and several other studies have demonstrated that inattention is an independent concurrent predictor of CDS in regression analyses [7, 8, 11, 15, 17, 18]. Together, these findings suggest that inattention and CDS often co-occur but having inattention at a young age does not necessarily predict later CDS, especially when potentially more relevant variables (e.g., autism and insomnia symptoms) are controlled.
Depression
Previous multivariate studies have shown that depression is a concurrent predictor of CDS in clinical and nonclinical samples [7, 8, 11]. Correlations between baseline depression and baseline CDS and between follow-up depression and follow-up CDS were medium to large in our study. Follow-up depression was a significant concurrent predictor of follow-up CDS. However, baseline depression did not significantly predict baseline CDS, and baseline depression was not a predictor of follow-up CDS. This may reflect the later onset of depression and the transient, episodic nature of depressive symptoms for some individuals. This would explain why the correlation between CDS and depression was medium to large both at young and older ages but not when depressive symptoms at a young age were compared with CDS symptoms 8 years later. Further, individuals who are depressed may be hypoactive and cognitively disengaged, which are core symptoms of CDS as well as symptoms of depression. In depressed individuals, CDS may represent a state more so than a trait.
Somatic Complaints
Somatic complaints (headaches, stomachaches, other body aches and pain, and complains of feeling sick) were significant concurrent predictors of CDS at both baseline and follow-up. The association between CDS and somatic complaints at a single point in time has been reported in other studies as well [7, 8, 11, 17]. In contrast, baseline somatic complaints did not predict follow-up CDS in our study when other variables were controlled. As with depression, these findings may reflect the transient nature of somatic symptoms for some individuals. Individuals who do not feel well may be hypoactive and cognitively disengaged, but they may not show these symptoms if and when they do feel well.
Anxiety
Although correlations between anxiety and CDS were medium at baseline and at follow-up, anxiety was not a concurrent or longitudinal predictor of CDS in our study when other variables were entered with anxiety in regression analyses. A review of multivariate studies in the literature reveals mixed findings. Anxiety was a significant concurrent predictor in some clinical and nonclinical samples [7, 11, 17], but not in an elementary school sample in another study [8]. More research is needed to determine what aspects of anxiety may or may not be related to CDS. Interestingly, one study [17] found that generalized anxiety, but not social and separation anxiety, were associated with CDS.
Externalizing Behavior Problems
Hyperactivity/impulsivity and ODD/conduct problems were nonsignificant concurrent and longitudinal predictors of CDS in our study, except that hyperactivity/impulsivity at baseline was a significant negative concurrent predictor of baseline CDS. Our findings are consistent with previous studies [7, 8, 11, 14, 17] showing the absence of a positive relationship between these externalizing problems and CDS.
Demographics
Demographics (age, sex, race, and parent occupation) were not related to CDS at baseline and at follow-up in our study. Published multivariate studies show that most demographics were nonsignificant concurrent predictors of CDS, including race [7, 13], sex [7, 15, 18], and age [7, 18]. Results for SES were mixed [7, 15, 18].
Psychological Test Scores
Only one multivariate longitudinal study predicting CDS was located in the literature [13]. This study used neuropsychological test scores and found that slower WPPSI-III Processing Speed at age 5 predicted CDS teacher ratings in 1st -3rd grade students. In contrast, our study found that WISC-III Processing Speed at a mean age of 9 years was not significantly correlated with CDS 8 years later. Further, correlations between CDS and all 15 IQ, achievement, and neuropsychological test scores at baseline were nonsignificant in our study. Only two of the baseline psychological tests scores (GDS Vigilance and Digit Span) were significantly correlated with the follow-up CDS score, but correlations were small (-.19 and − .17) and explained less than 4% of the variance. A review of the literature reveals relatively small or nonsignificant relationships between CDS and IQ, achievement, and neuropsychological test scores [11, 16, 18, 28, 55]. The absence of a strong relationship between CDS and both processing speed and reaction time is in part the SCT Work Group’s rationale for recommending that the term “sluggish cognitive tempo” be changed to CDS [1].