Trajectories of Insomnia Symptoms From Childhood Through Young Adulthood

The Penn State Child Cohort was designed as a random, population-based study of school-aged children for which details have been previously reported.^14,24–26  In brief, out of 5740 children (aged 5–12 years), 1000 of them were randomly selected to participate in an in-laboratory polysomnography (PSG) study (visit 1 [V1]) and 700 agreed (70% response rate) at baseline from 2000 to 2005. A total of 421 of these subjects underwent an in-laboratory follow-up study at ages 12 to 23 (visit 2 [V2]; 60.1% response rate) from 2010 to 2013.²⁵  No differences in demographic characteristics at V1 were observed in the 279 subjects lost to follow-up at V2.²⁵  The length between V1 and V2 in the 421 subjects ranged from 5.8 to 13 years (median = 7.4; interquartile range [IQR], 6.8–8.7). A total of 502 subjects (71.7% response rate since V1 and 79.1% since V2) aged 18 to 31 have completed a structured self-reported survey (survey 3 [S3]) from 2018 to 2021. No differences in demographic characteristics at V1 were observed between the 700 subjects and the 502 followed-up at S3, except a lower proportion of males (52.7% vs 48.0%). The length between V1 and S3 in the 502 subjects ranged from 11.4 to 20.3 years (median = 14.8; IQR, 13.4–16.3). The current study focuses on these 502 subjects with S3 data (n = 333 with data at all time points, n = 169 at V1 and S3) (Supplemental Fig 3).

Penn State College of Medicine’s institutional review board approved the study. Written informed consents from the parent or legal guardian and from subjects aged ≥18 years, and assent from those aged <18 years, were obtained.

At V1 and V2, subjects or their parents completed the Pediatric Behavior Scale (PBS)²⁷  and the Pediatric Sleep Questionnaire (PSQ).²⁸  At V1, 118 subjects were identified to have insomnia symptoms, defined as a parent report of “often/moderate” or “very often/severe” DIMS on the PBS^1,17,18,29  and/or sleep medication use (over-the-counter or prescription medications for DIMS and not for other medical or sleep issues)⁴ ; for 3 cases missing PBS data, a positive response to DIMS on the PSQ was used.¹⁴  At V2, 120 subjects were identified to have insomnia symptoms, defined as a self report of DIMS on the PSQ²⁸  and/or sleep medication use; for 13 cases missing PSQ data, a parent report of often/moderate or very often/severe DIMS on the PBS was used.^14,17,19  Only 1 subject was reclassified as having insomnia symptoms on the basis of their report of sleep medication use and mild DIMS. At S3, 214 subjects were identified as having insomnia symptoms, defined as a self report of “moderate” or “severe” DIMS and/or sleep medication use, on the basis of a structured questionnaire previously developed in the Penn State Adult Cohort.^5,6  In addition, 62 subjects who self-reported moderate or severe to the question “Do you have insomnia?” were identified as having adult insomnia at S3. This question has demonstrated adequate reliability and validity in the Penn State Adult Cohort^5,6  and further support for its validity in terms of moderate-to-severe daytime functioning impairment and greater nighttime sleep disturbance is shown in Supplemental Table 3. Thus, we use herein the term “insomnia symptoms” to refer to any presence of DIMS,^4–6  and use the term “adult insomnia” to refer to the subset of subjects who specifically reported having insomnia at S3.

At V1 and V2, sleep was monitored with in-laboratory PSG^1,24–26  from the time of “lights out” (21:00–23:00) until “lights on” (06:00–08:00) to adjust, as much as possible, to the subject’s habitual sleep schedule with 9 hours of time in bed, and was scored according to standard criteria.^30,31  Commensurate with previous studies, obstructive sleep apnea (OSA) was defined using pediatric criteria as an apnea hypopnea index of ≥2 per hour,^14,25,26  periodic limb movements (PLMS) as a PLM index of ≥5 per hour,^14,32,33  and OSSD as a total sleep time (TST) of <7.7 hours, the median of the cohort.^14,17,18  This cut-off remains consistent with our previous studies^14,17,18  and with other studies using actigraphy data, which provides a measure of habitual sleep duration.^34,35 

At S3, all 502 subjects self-reported on their TST for weekdays and weekends. Habitual sleep duration was calculated with the formula: [(weekdays TST × 5) + (weekends TST × 2)]/7 days. Self-reported habitual sleep duration at S3 in the 502 subjects ranged from 3.5 to 11 hours with a mean of 7.1 hours (median = 7.3; IQR, 6.6–7.8).

Subjects and parents completed a standard questionnaire with demographic information at V1, V2, and S3, with information including date of birth, sex, race, and ethnicity, as well as any psychiatric or learning disorders and psychotropic medication use.^24,25  At V1 and V2, height and weight were measured in the laboratory,^24,25,29  BMI percentile for sex and age was calculated using standard growth curves, and the presence of overweight was defined as a BMI percentile ≥85%. At S3, subjects reported on their height and weight, self-reported BMI was calculated using the standard formula, and the presence of overweight was defined as a BMI ≥25.

Descriptive statistics were used to report the demographic characteristics of the overall sample (Table 1). Our analytic approach examined the transitional probabilities of insomnia symptoms, their developmental trajectories and, thereafter, the associated risk of adult insomnia.

First, we examined the transitional probabilities of insomnia symptoms between each time point (ie, V1 to V2, V1 to S3, and V2 to S3) among the 502 subjects with V1 and S3 data and the 333 subjects with V1, V2, and S3 data.

Second, we estimated the developmental trajectories of insomnia symptoms among the 118 subjects with childhood insomnia symptoms (V1) and the developmental trajectories of normal sleep among the 384 subjects with normal sleep in childhood (V1). To fully use the data set, missing values for insomnia symptoms at V2 for 169 subjects were imputed on the basis of a logistic regression model built on the 333 subjects with complete data and under the missing-at-random assumption. The outcome of this regression model was the probability of insomnia symptoms at V2, and the predictors were insomnia symptoms at V1, insomnia symptoms at S3, and previously identified key risk factors (ie, sex, race and ethnicity, and psychiatric or learning disorders at V1).^1,14  Multiple imputations were performed and the SD and IQR among the 1000 imputations for each trajectory are reported; demographics and trajectories based on complete data can be found in Supplemental Table 4 and Supplemental Fig 4.

Third, we examined the percentage of subjects with normal sleep and insomnia symptoms at V1 or V2 who worsened into adult insomnia at S3. Logistic regression models were used to test if insomnia symptoms at V1 or at V2 were associated with increased odds of adult insomnia at S3 among subjects with or without OSSD at V1 or V2, on the basis of previous findings in youth and adults.^5,6,17–21  Covariates considered for inclusion for adjustment included sex, race and ethnicity, age (V1 or V2), overweight (V1 or V2), OSA (V1 or V2), PLMS (V1 or V2), psychiatric or learning disorder (V1 or V2), psychotropic medication use (V1 or V2), self-reported habitual sleep duration (S3), and time to follow-up (ie, from V1 to S3 or from V2 to S3). Stepwise regression procedures were used to select a subset for inclusion in the final model.

Analyses were conducted using SPSS version 26 and R version 4.1.0. P values ≤.05 were considered statistically significant.

As shown in Table 1, there were no significant differences in the sex, race, or ethnicity distribution between the 502 subjects with V1 and S3 data and the subset of 333 subjects who also had V2 data. The mean age at each time point was commensurate with the developmental stages of childhood (V1), adolescence (V2), and young adulthood (S3).

The prevalence of insomnia symptoms at V1 (23.5%) and at V2 (36.0%) did not significantly deviate from our previous cross-sectional studies,^1,14  whereas its prevalence at S3 (42.6%) was commensurate with that reported in other cross-sectional studies in young adults.² 

Table 2 presents the transitional probabilities of insomnia symptoms and normal sleep across all three time points. Among subjects with insomnia symptoms at V1, 53.7% persisted with insomnia symptoms at V2 and 61.9% did at S3, whereas 46.3% and 38.1% remitted during those periods. Among subjects with insomnia symptoms at V2, 57.5% persisted with insomnia symptoms at S3 and 42.5% remitted from insomnia symptoms at S3 (Table 2). A total of 44 subjects reported insomnia symptoms at both V1 and V2, of whom 81.8% persisted with insomnia symptoms at S3 and only 18.2% remitted by S3. Among the 38 subjects with insomnia symptoms at V1 who had remitted by V2, 60.5% remained remitted and 39.5% relapsed into insomnia symptoms at S3.

As shown in Fig 1A, among subjects with insomnia symptoms at V1 (n = 118), the most frequent predicted developmental trajectory was that of persistence (43.3% ± 1.6%, IQR, 42.4%–44.9% for 1000 imputations), followed by that of remission (26.9% ± 1.4%, IQR, 26.3%–28.0% since V1 and 11.2% ± 1.4%, IQR, 10.2%–11.9% since V2) and that of a waxing-and-waning pattern (18.6% ± 1.6%, IQR, 16.9%–19.5%).

Among subjects with normal sleep at V1, 69.7% remained with normal sleep at V2 and 63.3% did at S3, whereas 30.3% and 36.7% developed insomnia symptoms during those transitional periods. Similarly, among subjects with normal sleep at V2, 70.4% remained with normal sleep at S3 and 29.6% developed insomnia symptoms at S3 (Table 2). A total of 175 subjects reported normal sleep at both V1 and V2, of whom 72.6% continued to report normal sleep and 27.4% developed insomnia symptoms at S3. Among the 76 subjects with normal sleep at V1 who had developed insomnia symptoms at V2, 43.4% persisted with insomnia symptoms and 56.6% remitted into normal sleep at S3.

As shown in Fig 1B, among subjects with normal sleep at V1 (n = 384), the most frequent predicted developmental trajectory was remaining a normal sleeper (48.1% ± 0.9%, IQR, 47.7%–48.7%), followed by newly developing insomnia symptoms at S3 (20.7% ± 1.0%, IQR, 20.0%–21.3%), a waxing-and-waning pattern (16.0% ± 1.0%, IQR, 15.4%–16.7), and persisting with insomnia symptoms since V2 into S3 (15.2% ± 0.9%, IQR, 14.6%–15.6%).

The prevalence of adult insomnia at S3 was 12.3% (n = 62). A total of 22.0% and 20.8% of subjects with insomnia symptoms at V1 or at V2, respectively, reported adult insomnia at S3, whereas only 9.4% and 6.6% of subjects with normal sleep at V1 or at V2, respectively, reported adult insomnia at S3 (both P < .001).

As shown in Fig 2, multivariable-adjusted regression models indicated that insomnia symptoms at V1 were associated with 2.6-fold significantly increased odds of adult insomnia at S3 among those with OSSD at V1 (P = .039), whereas insomnia symptoms at V2 were associated with 5.5-fold significantly increased odds of adult insomnia at S3 among those with OSSD at V2 (P = .001). The odds of adult insomnia at S3 associated with insomnia symptoms among those with normal sleep duration either at V1 (P = .265) or at V2 (P = .749) were not significantly increased.

This is the first long-term cohort study to date to describe the developmental trajectories of childhood insomnia symptoms through young adulthood, as well as their evolution into adult insomnia by including PSG measures. The most common developmental trajectory for insomnia symptoms was that of persistence from childhood through young adulthood. Of those who persisted with childhood insomnia symptoms in the transition to adolescence, only 11% remitted by young adulthood. Additionally, nearly 20% of youth with insomnia symptoms worsened into adult insomnia, an incidence rate similar to adult studies with long follow-ups.^4,5  The odds of insomnia symptoms worsening into adult insomnia were 2.6-fold and 5.5-fold among children and adolescents, respectively, who slept objectively short in the laboratory, although were not significantly increased among those who had normal sleep duration. Thus, these 15-year longitudinal findings across 3 developmental stages indicate that insomnia symptoms should not be expected to developmentally remit in at least 40% of children and that adolescence is a critical developmental period for the adverse prognosis of the insomnia with short sleep duration phenotype.

Previous longitudinal studies examining insomnia symptoms across developmental stages have predominantly relied on self- and parent reports.^12,15,16  The current study used a combination of longitudinal parent- and self-reported data and in-laboratory PSG to determine the developmental trajectory of insomnia symptoms from childhood to young adulthood. Although the persistence of childhood insomnia symptoms has been reported at 30% to 56%,^14–16  a previous study found that remission, by early adolescence (26.2%) or by late adolescence (20.9%), was the most frequent trajectory of childhood insomnia symptoms, whereas 32.6% of children persisted with insomnia symptoms across adolescence.¹⁵  More recently, adolescent insomnia symptoms had a higher persistence rate (50%) in the transition to early adulthood.¹⁶  These seminal studies examined the transition from ages 7 to 9 through 11 to 13 into 16 to 19, restricting their findings to childhood and adolescence,¹⁵  or the transition from ages 16 to 18 to 22 to 25, limiting their findings to adolescence and young adulthood.¹⁶  The current study expanded upon this work by examining 3 developmental transitions and found that 27% of children with insomnia symptoms experience remission already in the transition to adolescence and confirms the high persistence (43%) and waxing-and-waning rates (18%) of insomnia symptoms using a within-subjects approach with longer follow-ups across these developmental stages.

Our novel data also indicated that PSG-measured short sleep, as typically assessed in sleep laboratories with fixed-time recordings, may be a useful assay for identifying youth with insomnia symptoms at greatest risk of worsening into adult insomnia, independent of OSA and PLMS. Previous studies have shown that youth who report insomnia symptoms and sleep objectively short differ from those with normal sleep duration in their association with neuroendocrine, inflammatory and cortical biomarkers, and adverse behavioral health outcomes.^17–21  However, we previously failed to demonstrate that children with OSSD were more likely to persist with insomnia symptoms in the transition to adolescence.¹⁴  Such finding suggested the persistence of childhood insomnia symptoms in the transition to adolescence is either mostly determined by behavioral factors^18,36–43  or by the strong homeostatic sleep drive present in children,⁴⁴  which may preclude the expression of increased biological vulnerability during that developmental period.¹⁴  The current study indicates that the odds of worsening into adult insomnia were greatest (5.5-fold) in adolescents who reported insomnia symptoms and slept objectively short, adding to previous data on their increased inflammation, cortical hyperarousal and depression risk,^19–21  and role of OSSD in the evolution of insomnia symptoms into insomnia disorder.⁵  These findings suggest that adolescence is a critical period for expression of the greater vulnerability underlying insomnia with short sleep duration, a more biologically severe phenotype associated with an unremitting course and cardiovascular and neurocognitive risk in adults.²²  Given that in-laboratory PSG is cumbersome and expensive, future studies should examine these associations using ambulatory PSG and/or actigraphy devices to obtain habitual estimates of sleep duration.^34,35 

This study has several limitations. First, measures of OSSD, OSA, and PLMS were collected via 1-night, 9-hour PSG, which may be affected by the first-night effect and may not be representative of habitual sleep timing and duration at home; however, our methods are translatable to current clinical practice where pediatric sleep studies are typically performed in-laboratory with fixed-time recordings. Second, the lack of PSG measures at S3 precluded controlling or examining the role of objective sleep and its trajectory into adulthood and only self-reported habitual sleep duration could be controlled for at S3. Third, while we collected data on the presence of a complaint of adult insomnia at S3 by relying on a previously validated epidemiologic method,^45–49  its presence as per strict diagnostic criteria could not be ascertained at any time point with the available self-reported data.

This study unveiled that the chronicity of childhood-onset insomnia symptoms is greater than previously believed, indicating that 43% of children with insomnia symptoms persist through young adulthood. Furthermore, this study revealed that 20% of children with insomnia symptoms worsen into adult insomnia, and that this increased risk is very high (5.5-fold) among adolescents who sleep objectively short. Pediatricians should consider adolescence a critical period for the adverse prognosis of insomnia symptoms, particularly for the insomnia with short sleep duration phenotype. Early sleep interventions are a priority, as clinicians should not expect insomnia symptoms to developmentally remit in a high proportion of children, and objective sleep measures may be indicated in adolescence to identify those with poorer long-term prognosis.

We thank the staff at the Clinical Research Center of Penn State College of Medicine, where this study was performed.

DIMS

difficulty initiating and/or maintaining sleep

IQR

interquartile range

OSA

obstructive sleep apnea

OSSD

objective short sleep duration

PBS

Pediatric Behavior Scale

PLMS

periodic limb movements

PSG

polysomnography

PSQ

pediatric sleep questionnaire

S3

survey 3

TST

total sleep time

V1

visit 1

V2

visit 2