|Home | About | Journals | Submit | Contact Us | Français|
Examine statistical effects of sleep-disordered breathing (SDB) symptom trajectories from 6 months to 7 years on subsequent behavior.
Parents in the Avon Longitudinal Study of Parents and Children reported on children’s snoring, mouth breathing, and witnessed apnea at ≥2 surveys at 6, 18, 30, 42, 57, and 69 months, and completed the Strengths and Difficulties Questionnaire at 4 (n = 9140) and 7 (n = 8098) years. Cluster analysis produced 5 “Early” (6–42 months) and “Later” (6–69 months) symptom trajectories (“clusters”). Adverse behavioral outcomes were defined by top 10th percentiles on Strengths and Difficulties Questionnaire total and subscales, at 4 and 7 years, in multivariable logistic regression models.
The SDB clusters predicted ≈20% to 100% increased odds of problematic behavior, controlling for 15 potential confounders. Early trajectories predicted problematic behavior at 7 years equally well as at 4 years. In Later trajectories, the “Worst Case” cluster, with peak symptoms at 30 months that abated thereafter, nonetheless at 7 years predicted hyperactivity (1.85 [1.30–2.63]), and conduct (1.60 [1.18–2.16]) and peer difficulties (1.37 [1.04–1.80]), whereas a “Later Symptom” cluster predicted emotional difficulties (1.65 [1.21–2.07]) and hyperactivity (1.88 [1.42–2.49]) . The 2 clusters with peak symptoms before 18 months that resolve thereafter still predicted 40% to 50% increased odds of behavior problems at 7 years.
In this large, population-based, longitudinal study, early-life SDB symptoms had strong, persistent statistical effects on subsequent behavior in childhood. Findings suggest that SDB symptoms may require attention as early as the first year of life.
Sleep-disordered breathing is associated with neurobehavioral morbidity in children. Prior related research has generally been cross-sectional or short (ie, 1–2 years) follow-up studies of a single symptom (ie, snoring, obstructive sleep apnea, mouth breathing), with limited control for confounders.
Sleep-disordered breathing was assessed as a trajectory of combined symptoms from 6 months to 69 months, in more than 11000 children. Sleep-disordered breathing was associated with 40% and 60% more behavioral difficulties at 4 and 7 years, respectively.
Neurobehavioral morbidity is common in childhood sleep-disordered breathing (SDB) that can range from snoring to obstructive sleep apnea. Mouth breathing is another frequent clinical finding.1,2 SDB causes abnormal gas exchange, interferes with sleep’s restorative processes, and disrupts cellular and chemical homeostasis.3 The supposed resultant dysfunction of the prefrontal cortex impairs attention, executive functioning, behavioral inhibition, self-regulation of affect and arousal, and other socio-emotional behaviors.4 Behavioral manifestations include both externalizing (eg, hyperactivity, aggression, impulsivity) and internalizing (eg, somatic complaints, social withdrawal) behaviors.5 SDB reportedly peaks from 2 to 6 years of age,6 but also occurs in younger children.7 SDB’s neurologic effects may be irreversible,8 highlighting the saliency of underdetection.
SDB presents as a heterogeneous disorder in children. Understanding how and when SDB symptom patterns in early life affect neurobehavioral outcomes has clinical implications for deciding whether, how, and in whom to intervene.9 Yet, existing studies of SDB’s neurobehavioral effects in children are primarily cross-sectional, and limited by poor sampling, insufficient consideration of confounders, and imprecise use of statistical tools.10,11 The few longitudinal studies are either before or after tonsillectomy or follow children for ≤2 years.
This study describes the combined trajectory of 3 hallmark SDB symptoms (snoring, mouth breathing, and witnessed apnea) and their longitudinal statistical effects on behavior. Our research questions were (1) What effect do early SDB trajectories, from 6 through 42 months of life, have on social-emotional behavior at 4 and 7 years? and (2) What effect do SDB trajectories from 6 months through 69 months have on behavior at 7 years of age? We analyzed previously collected observational data from a critical period in SDB development, from 6 months through nearly 7 years of age in a prospective, population-based cohort.
The Avon Longitudinal Study of Parents and Children (ALSPAC), a geographically based cohort study of children, enrolled pregnant women residing in a defined part of the former county of Avon in southwest England with an expected date of delivery between April 1991 and December 1992. A total of 14541 pregnant women were enrolled. The cohort, described in detail elsewhere,12 is broadly representative of the UK population in terms of socioeconomic status (SES), although with a slight underrepresentation of ethnic minority families, and overrepresentation of wealthier families. Our analyses excluded twin, triplet, and quadruplet births; children who did not survive to 1 year; and children with conditions, such as major congenital disorders, that are likely to affect SDB or behavioral assessment. The resulting base sample, used to derive SDB clusters and behavioral outcomes, was 13467infants.
ALSPAC’s internal law and ethics committee reviews all proposals for secondary analyses and approves policies for data handling and analysis. Ethical approval for this analysis was obtained from the ALSPAC Law and Ethics Committee and UK Local Research Ethics Committee. All participants provided informed consent.
Questionnaires, designed and mailed as part of the original ALSPAC study when children were 6, 18, 30, 42, 57, and 69 months of age, asked parents to report on their child’s snoring, apnea, and mouth breathing. These items were as follows: (1) Mouth breathing: “Does he or she breathe through the mouth rather than the nose?” At 57 months and older, parents were asked to report separately for mouth breathing when awake versus asleep, although only the latter was used in analyses. (2) Snoring: “Does he or she snore for more than a few minutes at a time?” (3) Apnea: “When asleep, does he or she seem to stop breathing or hold breath for several seconds at a time?” The ALSPAC parent-reported SDB measures are similar or identical to items validated against polysomnographic data from sleep laboratories. Some validated questionnaires have included parent report of all 3 SDB symptoms,13–17 whereas others have included only snoring and apnea.18,19
Responses were categorized along ordinal scales of 3, 4, or 5 levels. Given this inconsistency in (preexisting) response categories, we extrapolated the values to a common scale (0–100) with the “Always” responses anchored at one end and the “Never” or “Rarely/Never” responses anchored at the other, and proportionate spacing in-between (ie, a 4-category scale was recoded as 0, 33, 66, 100). Variables were then transformed to z scores, with higher scores indicating greater symptom burden.
The Strengths and Difficulties Questionnaire (SDQ),20 a widely used behavioral screen, was completed by mothers when children were ~4 and 7 years old. The 25-item SDQ has 5 scales: inattention/hyperactivity, emotional symptoms (anxiety and depression), peer problems, conduct problems (aggressiveness and rule breaking), and a pro-social scale (sharing, helpfulness, and so forth). A total difficulties (range = 0–40) score is generated by summing all but the latter scale because the absence of pro-social behavior is conceptually different from the presence of psychological difficulties. Higher scores denote more problems. Missing data were prorated according to SDQ instructions.21 The SDQ scores were dichotomized at the upper 10% based on psychometric testing,22 ALSPAC,23,24 and other UK cohort studies.25
Initial covariate selection was guided by previous ALSPAC studies of SDQ outcomes,24,26–32 and non-ALSPAC studies of sleep problem effects on SDQ outcomes.33,34 Based on this literature review, and exploratory analyses, 15 potential confounders were incorporated into analyses.
SES was measured by paternal employment (manual versus professional), maternal education (higher versus lower), and housing inadequacy (if either >1 person/room or homeless). Family adversity was measured by an 18-item index of stressors (eg, maternal psychopathology, crime, financial insecurity) used in other ALSPAC analyses35; higher values signify more adversity. Intrauterine exposures of maternal smoking or alcohol use in the first trimester (yes/no) and fish intake at 32 weeks’ gestation (servings/week) were assessed, as was whether the child was ever breastfed and the mother’s age at delivery. Household variables included family size (0, 1, or ≥2 children in household at 6-month interview) and the Home Observation for Measurement of the Environment (HOME) Inventory36 to assess the quality of parenting and home environment. Child demographics included race (white versus other) and gender, low birth weight (<2500 g) and prematurity (<37 weeks). Analyses of the subsample with BMI z score data are in Supplemental Tables.
To predict behavior at 4 years, “Early” clusters were derived from the 11049 participants in the base sample with SDB data for ≥2 of the first 4 time points (ie, 6, 18, 30, and/or 42 months). To predict behavior at 7 years, “Later” clusters were derived for the 11235 participants in the base sample with SDB data for ≥2 of the first 6 time points (ie, 6, 18, 30, 42, 57, and/or 69 months). Through a process described in detail elsewhere,37 we produced 5 conceptually and statistically distinct Early clusters (6–42 months), and 5 comparable Later clusters (6–69 months) that were extensions of Early clusters.
For SDQ scores at 4 and 7 years, we calculated the mean (SD) and proportions above and below the 10% cutoff for the base sample, and their associations with putative covariates either from χ2 test or analysis of variance. We describe the association between SDQ mean (SD) total scores at 4 and 7 years and the Early and Later clusters, by analysis of variance. Only participants not missing SDQ data are included in analyses of behavioral outcomes.
Multivariate logistic regression analyses examined adjusted and unadjusted relationships between clusters and SDQ total and subscales at 4 and 7 years. To streamline presentation, unadjusted analyses are in the Supplemental Tables. Initial models included all putative covariates. Only those variables that were significant (P ≤ .05) were retained in multivariate models. Odds ratios (ORs) and 95% confidence intervals (95% CIs) represent the odds of being in the top 10% versus the remaining 90% of SDQ scores. To address multicollinearity, variance inflation factors were derived to assess the effects of individual independent variables on variance. A conservative variance inflation factor threshold of 10 was used in model testing.38 Analyses were conducted by using SAS version 9.1 (SAS Institute, Inc, Cary, NC).
SDB longitudinal data were relatively complete. Early cluster analyses of the 7996 participants with SDQ 7-year outcomes included 7716 (96%) with SDB data for ≥3/4 time points. Likewise, Later cluster analyses (SDQ 7-year data) for 8064 participants, included 7383 (92%) with SDB data for ≥5/6 time points. Missing SDQ or SDB data were significantly associated with nonwhite race, prematurity, low birth weight, manual (versus professional) paternal employment, lower (versus higher) maternal educational status, housing inadequacy, not being breastfed, and higher levels of wheezing (not shown).
Characteristics of the base sample and associations with behavioral outcomes are shown in Table 1. Children in the upper 10% of SDQ scores had significantly more adverse characteristics (eg, higher maternal smoking, older delivery age, lower maternal education, higher Family Adversity Index scores, lower HOME scores, housing inadequate, prematurity, low birth weight, and being male) than the remaining 90% at 4 and 7 years, but there were no differences by race, maternal alcohol intake during pregnancy, or tonsils removal.
Cluster analyses yielded 1 asymptomatic (“Normals,” 45% of sample) and 4 symptomatic (55% of sample) trajectories. Early clusters, shown in Fig 1, can be summarized as (1) symptoms “Peak @ 6” and then abate, (2) symptoms “Peak @ 18” months and then abate, (3) symptoms peak at 30 months and then persist (“Worst Case”), (4) symptoms emerge at 42 months and then persist (“Late Symptom”) and (5) “Normals” who are asymptomatic throughout. Snoring levels were nearly double those of apnea or mouth breathing in Worst Case and Late Symptom versus comparable symptom levels in other clusters. Whether assessed as a continuous or dichotomous (10% vs 90%) variable, SDQ total scores differed significantly across the 4 symptomatic clusters, and in combined symptomatic clusters versus Normals (Table 2). Early clusters’ descriptive characteristics are shown in Supplemental Table 6.
Five comparable Later clusters are illustrated in Fig 2. Patterns are similar to the Early clusters, except that in this Late Symptom cluster, snoring and mouth breathing peak together at lower levels at 57 months with no marked apnea, and the Peak @ 6 apnea levels are nearly double those of the Early clusters.
The SDB clusters significantly predict SDQ total scores at 4 and 7 years (Table 3). Early cluster effects are 20% to 70% in multivariate analyses. The strongest and most persistent is for Worst Case, with comparable outcomes at 4 (OR = 1.49, 95% CI = 1.11–1.99) and 7 years (OR = 1.72, 95% CI = 1.31–2.25). Later clusters’ effects are 40% to 100% in multivariate analyses, including a ≈40% effect for Peak @ 6 and ≈50% effect for Peak @ 18. Membership in any symptomatic cluster is associated with being in the upper 10% of total SDQ scores (versus Normals), an effect that increases slightly from 4 (OR = 1.33, 95% CI = 1.14–1.56) to 7 years (OR = 1.52, 95% CI = 1.24–1.85: Later clusters). Multivariate effects are strongly attenuated from unadjusted effects (Supplemental Table 7).
SDB effects exceeded those of maternal smoking, maternal education, and paternal employment in multivariate analyses (Table 3). Race, prematurity, low birth weight, maternal fish intake, or maternal age were not significant in any multivariate analyses, total or subscale (not shown). Only male gender and not being the second or later-born child had greater adverse effects. Greater family adversity was associated with poorer behavioral outcomes, whereas higher home environment scores were associated with improved outcomes. Gender-cluster interactions were not significant for any total or subscore analyses.
In adjusted analyses, nearly every cluster-subscale association was significant (Table 4), with effects of 20% to 100% (see Supplemental Table 7 for unadjusted effects.) Details are in the following paragraphs.
Peak @ 6 was associated with ≈30% greater odds of being in the lowest decile across outcomes at 4 and 7 years. All of the remaining associations were not significant.
With 1 exception, all effects (≈20%–100%) were significant, and increased from 4 to 7 years. Furthermore, Early cluster effects at 4 years for Worst Case and Late Symptom equaled or increased at 7 years.
With 1 exception at 4 years, all effects were significant (range ≈20%–65%), and most increased from 4 to 7 years. Late Symptom had the strongest effect at 4 and 7 years based on Early cluster models, with effects persisting to 7 years in Later cluster models.
With 1 exception, all effects were significant (range ≈30%–70%). Both Worst Case and Peak@18 effects increased from 4 to 7 years, whereas Late Symptom and Peak @ 6 effects attenuated over that time.
Half of the cluster-subscale associations were significant, and SDB effects were more modest (range ≈30%–50%) and stable over time, compared with the other subscales. Effects were strongest for Worst Case.
We examined the effects of snoring, apnea, and mouth-breathing patterns (clusters) on behavior, from infancy through 7 years in more than 11000 children. By 4 years, children in the symptomatic clusters were ≈20% to 60% more likely to exhibit behavioral difficulties consistent with a clinical diagnosis; by 7 years, they were ≈40% to 100% more likely. These effects, in a population-based cohort that controlled for 15 putative confounders, exceeded those of any measured prenatal (ie, maternal age, smoking or alcohol use), gestational age, birth weight, breastfeeding, SES, family adversity, or home environment exposure. Furthermore, SDB effects at 4 years were as predictive of behavioral difficulties at 7 years. The worst symptoms were associated with the worst behavioral outcomes. Among the neuroehavioral domains assessed, hyperactivity was most affected.
Compared with previous parent-reported effects of SDB on Later behavior, our findings are conservative. In a study of ≈1000 third graders, snoring at baseline was associated with 2- to 10-fold increases in SDQ-assessed hyperactivity, and emotional, conduct, and peer difficulties at 1-year follow-up in age- and gender-adjusted analyses.39 A pediatric clinic sample of 229 2- to 13-year-olds found that baseline SDB symptoms predicted fourfold increases in hyperactive behaviors at 4-year follow-up, after adjustment for age, gender, and baseline hyperactivity.40 These studies differed from ours, with smaller, less representative samples, lack of data from the earliest years, use of other (non-SDQ) behavioral measures, and limited control for confounders. Alternatively, several cross-sectional studies that used different operational definitions of SDB and SDQ outcomes found no effects. One, a large nationally representative cross-sectional sample of ≈5000 Australian 4- to 5-year-olds, found that neither SDQ total, nor the Hyperactivity, Peer, or Emotional scale (actual) scores were greater among children with “snoring and/or breathing difficulties” during sleep ≥4 times per week.33 Likewise, among 635 6- to 8-year-olds, snoring ≥1 time per week in the past 6 months was not associated with high (upper 10th percentile) SDQ scores.34
This is the first study to assess SDB as a trajectory of combined symptoms, across a key period of SDB development from 6 months to 81 months, in a large sample. Previous studies had smaller samples that were often cross-sectional, or had shorter follow-up. Many were not population-based, involved school-age children only, or did not adjust for as wide a range of confounders.8,41 The potential impact of confounders is illustrated by the fact that several covariates independently associated with the clusters were not significant in multivariate analyses, and most unadjusted effects of SDB attenuated in controlled analyses. Although residual confounding is possible, covariates were selected based on previous ALSPAC analyses of SDQ outcomes, and non-ALSPAC studies of sleep problem effects on SDQ outcomes.
The current study has several limitations. First, SDB data were derived from parent report, rather than objective testing; however, the symptom-items used reflect widely accepted and well-validated SDB risk factors. Second, because parents may hear snoring from another room, they may be more likely to report it than either apnea or mouth breathing. Third, observed apnea during infancy is difficult to distinguish from the more common central apnea; however, our finding that observed apnea in infancy tracked with more clearly obstructive symptoms through 7 years suggests that observed apneas of infancy may not always have a central etiology. Fourth, missing SDB and SDQ data were associated with identified SDB risk factors (eg, maternal smoking, lower SES). Although biases that involve selective dropout may alter prevalence estimates, other ALSPAC analyses found only marginal effects on regression models predicting behavioral outcomes.42 This is likely the case in our study, in which such biases would render our findings more conservative.
These findings, from the largest-ever cohort study of SDB exposure and neurobehavioral morbidity, provide epidemiologic evidence that early childhood SDB effects may only become apparent years later. The most significant long-term effects occurred in children with the greatest overall levels of snoring, apnea, and mouth breathing throughout, peaking at 30 months. Even very early symptom peaks at 6 and 18 months are associated with ≈40% and 50%, respectively, increased behavioral morbidity at 7 years of age. This may be because of the increased vulnerability to SDB effects during this early critical period of brain development,43 when there is the greatest need for sleep.44
SDB is relatively common in childhood. In previous analyses of this cohort, the prevalence of habitual snoring ranged from 10% to 21%, from 6 months to 81 months.45 The potential clinical and educational implications of untreated SDB, therefore, are notable. As an example, we found significant (nearly twofold), and sustained effects on hyperactivity. In national survey data, children with attention-deficit/hyperactivity disorder had increased adjusted risks of comorbid learning disability (eightfold), anxiety (eightfold), and low social competence (threefold).46 Further, 40% to 80% of the nation’s 3 million 6- to 21-year-olds who receive special education for a developmental disability or delay also have attention-deficit/hyperactivity.47
These population-based data found a strong and persistent association between SDB symptoms and behavior. This has clinical implications for screening and treatment. A 2009 consensus statement by UK pediatricians and pediatric specialists noted that “the natural history of SDB, where a child changes from normality to abnormality, and where the risks of developing complications of the condition outweigh the risks of the surgical intervention, has not been established.”48 Although data from multicenter, randomized controlled trials, such as the current National Institutes of Health–funded Childhood Adenotonsillectomy study, will provide some evidence of cause-and-effect relationships, our findings provide further epidemiologic evidence to support attention to SDB symptoms beginning as early as the first year of life.
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting the families, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. In addition, we thank the following members of the study’s advisory group: Dr Raanan Arens (Montefiore Medical Center/Albert Einstein College of Medicine), Dr John Bent (Montefiore Medical Center/Albert Einstein College of Medicine), Dr Peter Blair (University of Bristol), Dr Pauline Emmett (University of Bristol), Dr Peter Fleming (University of Bristol), Dr Jon Heron (University of Bristol), Dr Carole Marcus (Children’s Hospital of Philadelphia), Dr Kenneth Ong (Cambridge University), Dr Sanjay Parikh (Montefiore Medical Center/Albert Einstein College of Medicine), and Dr Susan Redline (Case Western Reserve University).
All authors meet the criteria for authorship; Dr Bonuck conceptualized and designed the study, drafted the initial manuscript, reviewed and modified the analyses in collaboration with Drs Freeman and Xu, and incorporated coauthor feedback into the final manuscript; Dr Freeman worked to develop the methods, carried out initial analyses, supervised final analyses of Dr Xu, and reviewed and revised the final manuscript; Dr Chervin advised on study design and analyses, and carefully reviewed and revised multiple versions of the manuscript; Dr Xu collaborated on statistical design issues, completed final analyses, and reviewed and revised the final version of the article.
Dr Xu's current affiliation is Department of Pediatrics, Baylor College of Medicine. Dr Freeman is currently semi-retired.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by grants from the National Heart, Lung, and Blood Institute (R21HL091241 and R21HL091241-01A1). Funded by the National Institutes of Health (NIH).