Variation of Cognition and Achievement With Sleep-Disordered Breathing in Full-term and Preterm Children

Objective  Pediatric sleep-disordered breathing (SDB) has a disproportionately high prevalence in children who were preterm infants (preterm children) and is associated with behavioral comorbidity. Exposure to intermittent hypoxemia and sleep disruption may contribute to cognitive impairment. We quantified the association of SDB with cognition and achievement and determined whether preterm children are at a differentially increased risk for SDB-related impairments.

Design  Cross-sectional analyses.

Setting  Urban community.

Participants  Eight hundred thirty-five children, aged 8 to 11 years.

Intervention(s)  None.

Main Outcome Measures  Scores on the Peabody Picture Vocabulary Test–Revised, Kaufman Assessment Battery for Children, and Continuous Performance Test.

Results  One hundred sixty-four children had SDB. In unadjusted analyses, children with SDB had poorer scores on almost all tests of cognition and achievement. Group differences were attenuated after adjusting for socioeconomic status; in these analyses, children with SDB scored lower on the Peabody Picture Vocabulary Test–Revised (mean ± SE score, 100.5 ± 1.4 vs 103.6 ± 0.7; P = .04), and the Kaufman Assessment Battery for Children riddles and triangles subscales. Associations were stronger in preterm than in full-term children. Of the sleep measures, snoring history was most strongly correlated with indices of cognition and achievement.

Conclusions  Although moderate group differences were observed for almost all cognitive measures, an attenuation of effects was observed once socioeconomic status was considered. The deficits in selective measures of academic abilities, language comprehension, and planning and organizational skills suggest a negative impact of SDB on achievement and cognition. Stronger associations in preterm children suggest the importance of screening for snoring, a good predictor of cognitive function in this population.

Sleep-disordered breathing (SDB), a common but underrecognized chronic health problem, is characterized by episodic periods of occlusion of the upper airway during sleep, often associated with oxygen desaturation, arousal, and impaired sleep quality. Symptoms include loud snoring, frequent awakenings, and excessive sleepiness. In adults, SDB is associated with a myriad of neurocognitive difficulties, including deficits in general intellectual function, attention, memory, and organizational abilities.^1-3

Pediatric SDB may affect 2% to 4% of children, with a higher prevalence in certain subgroups, such as African American children and those who were preterm infants (preterm children).⁴ The health impact of childhood SDB, however, is not well understood. Because learning is incremental and a strong foundation is essential during childhood, it is important to evaluate the association of SDB with children's cognitive and academic skills. Previous research has been limited by small numbers of participants,^5,6 a focus on children referred for clinical evaluation (in whom referral may have been stimulated by behavioral or cognitive concerns),^5,6 nonstandardized assessments of cognition,⁷ exclusive reliance on parental observations of SDB symptoms,⁸ incomplete control for confounding factors,⁹ and inclusion of small numbers of children with objective evidence of sleep apnea.¹⁰

This study evaluates the performance of 835 children from a community-based sample on objective assessments of cognitive skills and academic achievement. We compared subgroups with and without SDB to address the following hypotheses: (1) SDB is associated with deficits in cognitive abilities and achievement; (2) preterm children, because of increased risk of neurodevelopmental disabilities related to early-life hypoxemia and ischemic injury, are at a differentially increased risk for SDB-related cognitive dysfunction; and (3) severity of overnight hypoxemia will be positively correlated with the extent of cognitive deficits.

Methods of recruitment to the Cleveland Children's Sleep and Health Study have been previously reported.⁴ In brief, the Cleveland Children's Sleep and Health Study is an urban community-based cohort of 907 children studied at 8 to 11 years of age. It was originally assembled as a stratified random sample of full-term and preterm children born at hospitals in the Cleveland, Ohio, area between January 1, 1988, and December 31, 1993. This cohort was designed to overrepresent African American and preterm children. Participants were excluded from analyses owing to inability to ascertain SDB status (n = 65) or missing cognitive function data (n = 7).

Children underwent assessment at 8 to 11 years of age during a home visit, when health information was ascertained with the use of a pediatric modification of a validated questionnaire.¹¹ Trained research assistants (J.L.E. and A.M.Z.), blinded to the child's SDB status, administered the cognitive tests. On the night after the testing battery, children underwent an in-home sleep study. A small proportion (14%) of families and children were interviewed in a research setting rather than the home.

Institutional review boards at the participating hospitals approved the protocol. Written informed consent was obtained from the child's guardian, and assent was obtained from the child.

Details on the methods for the assessing SDB have been reported.⁴ In brief, a 6-channel monitor recorded respiration (with rib cage and abdominal inductance plethysmography), heart rate, oxygen saturation level, and body position (SomnoStar PT-2; SensorMedics, Yorba Linda, Calif). Obstructive sleep apnea was defined as an obstructive apnea-hypopnea index (OAHI) of greater than or equal to 5 events per hour and/or an obstructive apnea index (OAI) of greater than or equal to 1 event per hour. The approach for quantifying obstructive sleep apnea has been validated by comparing the OAHI from the home study with a similar index from findings of 12-channel laboratory-based polysomnography in 55 children in whom the 2 tests were performed within 3 months (intraclass correlation coefficient, 0.85).⁴ Snoring was ascertained via a questionnaire that requested the parent to report the frequency of loud snoring by the child during the past month on a 5-point scale, ranging from never to always. Habitual snoring was defined as loud snoring during the past month, occurring at least 1 to 2 times per week. Children with obstructive sleep apnea and/or who were habitual snorers were categorized as having SDB, whereas children who were not habitual snorers and did not have obstructive sleep apnea were categorized as not having SDB. For exploratory analyses, children were also classified as nonsnorers vs snorers (ie, occurring rarely or more often per week, during the past month).

Cognitive skills and academic achievement were assessed with the Peabody Picture Vocabulary Test–Revised (PPVT-R),¹² Kaufman Assessment Battery for Children (K-ABC),¹³ and Continuous Performance Test (CPT).¹⁴ To administer the PPVT-R, spoken words are presented in increasing order of difficulty, and the examinee is asked to point to 1 of 4 pictures that shows the meaning of the word. The test measures comprehension of spoken words but also requires the ability to selectively attend to the response choices. The K-ABC consists of 13 subscales and 5 composite scales assessing a range of mental processing abilities and written language and mathematics skills. The K-ABC mental processing composite consists of sequential and simultaneous global mental skills. Specific subscales require memory, eye-hand coordination, spatial skills, problem solving, and executive functions that include the abilities to attend to the task and plan and organize one's approach. The CPT is a 14-minute computerized test of sustained attention and response inhibition. This test requires the child to press a key whenever an x appears on the screen and not to press the key when nontarget stimuli are presented. Time to respond, correct and incorrect responses, and missed targets are recorded. Raw scores on the K-ABC and PPVT-R were converted to age-adjusted standard scores based on published normative values.^13,14 Standard scores for the PPVT-R and K-ABC global measures have a mean of 100 and SD of 15, with higher scores reflecting better cognitive performance. Our primary outcome measures were standard scores from the PPVT-R, the K-ABC mental processing composite and achievement scales, the reaction time, and an index describing the child's ability to detect target stimuli while inhibiting responses to nontargets (response discrimination, D-prime) from the CPT.

Analyses were designed to obtain an unbiased estimate of the association between SDB and each measure of cognitive performance. Covariates and potential confounders, including the child's age, sex, birth weight, and indices of socioeconomic status (SES), were selected on the basis of theory and empirical research. Socioeconomic status was assessed using data on caregiver education, caregiver marital status, and census tract median income. Median income from the 2000 census was obtained by matching the child's address to the corresponding census tract. To control for prematurity and birth weight, the following 4 mutually exclusive categories were created on the basis of a combination of preterm status and birth weight: premature with extremely low birth weight (<1000 g); premature with very low birth weight (1000-1499 g); premature with low birth weight (≥1500 g), and full-term with normal birth weight (≥2500 g).

Groups were compared using 2-sample t tests, Wilcoxon rank sum tests, and the Pearson χ² test for normally distributed, skewed, and categorical data, respectively (SAS software, version 9.1; SAS Institute Inc, Cary, NC). Spearman correlations were used to assess the magnitude of the association between cognitive outcomes and sleep measures. Multiple linear regression was used to assess the relationship between SDB and each cognitive outcome, adjusting for potential confounders. Graphical techniques were used to assess the relationship between continuous covariates and cognitive outcomes. Models were fit without covariate adjustment (unadjusted); adjustment for age, sex, and birth weight category (partially adjusted); and adjustment for age, sex, birth weight category, caregiver education and marital status, and median income (fully adjusted). Unless otherwise indicated, data are expressed as mean ± SD.

The sample was approximately 50% female and had a mean age of 9.5 ± 0.8 years (Table 1). By design, approximately half of the sample (45.9%) consisted of preterm infants, including 9.0% with extremely low birth weight (<1000 g). The median income from the census tract data was $46 000 per year, and the median poverty rate was 6.2%.

One hundred sixty-four children (19.6%) children were classified as having SDB. As reported previously,⁴ children with SDB were significantly more likely to be African American and to have been born prematurely. They also were more likely to live in a neighborhood with lower median income, live in a single parent household, and have a caregiver with a high school education or less. The 2 groups did not significantly differ with respect to age and sex. Tonsillectomy and/or adenoidectomy were reported more commonly among children with SDB compared with children without SDB.

Sleep characteristics of the sample are shown in Table 2. Overall, 21 children had an OAHI of 5 or greater, and 34 children had an OAI of 1 or greater. As expected, the percentage of sleep time at an oxygen desaturation level of less than 90% was higher among children with SDB. Approximately two thirds of parents reported that their child never snored, whereas 17.6% reported habitual snoring. Parents reported that their child slept 9.2 ± 0.9 hours on average, with no difference for children with and without SDB.

Table 3 shows the distribution of the unadjusted, partially adjusted, and fully adjusted PPVT-R and K-ABC scores. The unadjusted and partially adjusted (ie, not including indices of SES) results show that the SDB group scored significantly lower on the PPVT-R and K-ABC summary measures. The unadjusted analyses showed an average difference of 9 points for the PPVT-R and 5 points for the K-ABC summary scales, effect sizes that are moderate to large. There was a small attenuation of group differences once the effects of age, sex, and birth weight were considered (partially adjusted results). Further adjustment for SES, with consideration of caregiver education and marital status and median income (the latter explaining the largest proportion of variance in cognitive measures), resulted in considerable attenuation of SDB group differences. Although in fully adjusted analyses the SDB group tended to score lower on all tests and subtests, significantly lower scores were observed for the PPVT-R and the K-ABC riddles and triangles subscales only. Performance measured during the CPT showed a longer mean reaction time in children with SDB; however, this measure and response discrimination were not significantly associated with SDB after covariate adjustment. Excluding children who had had a tonsillectomy and/or adenoidectomy reduced statistical power slightly, but provided qualitatively similar results.

To address whether associations of SDB with cognitive and achievement skills may be stronger in a vulnerable group, analyses were repeated stratifying by preterm status, and the results for the preterm group are presented in Table 4. The fully adjusted models showed that preterm children with SDB performed significantly worse on the PPVT-R and the K-ABC achievement and simultaneous global scales and riddles and triangles subscales. In contrast, the differences in neurocognitive function between the full-term children with and without SDB did not achieve statistical significance after fully adjusting for covariates (data not shown).

To explore which physiological stresses associated with SDB are most closely associated with neurocognitive deficits, the strength of the associations of overnight oxygen desaturation; frequencies of apnea and hypopnea and snoring; and sleep time with each cognitive outcome were assessed. The unadjusted and fully adjusted correlations show weak associations between cognitive outcomes and the oxygen saturation measures and OAHI and OAI variables, particularly after covariate adjustment (Table 5). Sleep time shows a modest, positive correlation with cognitive measures in unadjusted analyses only. Frequency of snoring has the strongest association with the cognitive outcomes, which are attenuated but remain statistically significant after covariate adjustment.

Graphic techniques used to further assess the association between snoring frequency and cognitive performance showed that children who were reported to never snore in the preceding month had higher scores compared with snoring children, with no difference in those in whom snoring frequency was reported as rare compared with more frequent. Analyses that substituted snoring status as the primary predictor rather than SDB showed that the mean differences between groups were slightly greater compared with the main analysis in which SDB was the primary predictor (Table 6). In fully adjusted analyses, many of these group differences are statistically significant, with snorers demonstrating lower scores on the PPVT-R, K-ABC achievement global scale, and K-ABC subscales arithmetic (97.8 ± 0.9 vs 100.7 ± 0.7; P = .007), faces and places (100.8 ± 0.9 vs 103.2 ± 0.6; P = .01), reading/decoding (99.1 ± 0.8 vs 101.9 ± 0.6; P = .005), reading/understanding (100.8 ± 0.8 vs 103.0 ± 0.6; P = .02), and riddles and triangles (Table 6). Significant differences in the CPT measures were not observed. Results were essentially unchanged in analyses excluding children with an elevated OAHI or OAI.

A major component of SDB morbidity is thought to relate to neuropsychological deficits, with concern that in children, SDB may result in long-term decrements in academic achievement.⁸ Although childhood SDB has been associated consistently with behavioral morbidity, including conduct problems and aggressiveness,^15-18 fewer data are available that address the association of childhood SDB and cognitive performance and achievement. This study extends previous research, addressing the association of a range of cognitive functions with SDB in a large, diverse pediatric population. The availability of several measures of SES allowed us to rigorously address potential environmental confounders. Inclusion of a large preterm group permitted assessment of low birth weight and prematurity as possible confounders while also providing an opportunity to explore differential susceptibility of a preterm population. These findings show (1) moderate to large SDB group differences for almost all cognitive measures (ie, effect sizes of 0.3-0.6), but evidence of considerable attenuation of group differences once SES (particularly neighborhood median income) was considered; (2) persistence of small to modest group differences for the PPVT-R and 2 K-ABC subscales (riddles and triangles) in adjusted analyses; (3) stronger associations of SDB with deficits in cognitive ability and academic achievement in preterm compared with full-term children; and (4) evidence that of all SDB measures (including oxygen saturation, the frequency of apneas and hypopneas, and sleep time), snoring history was most strongly associated with cognitive deficits.

Although large deficits in a broad range of neurocognitive tests were observed between SDB groups in unadjusted or partially adjusted analyses, differences were attenuated once SES was accounted for by adjusting for caregiver education and marital status and neighborhood income level. To our knowledge, this is the first time that an index of neighborhood poverty has been considered in relation to SDB. Because this measure may serve as a proxy for a wide variety of social and environmental risk factors that influence development, including quality of education and neighborhood structure, it may provide additional information over what is usually available from parent self-reported income. The findings suggest that some of the effects of childhood SDB in the literature may have been overestimated owing to incomplete consideration of confounding due to SES and premature birth or owing to other selection biases, eg, studying children at high risk for problems in behavior, cognition, and achievement.

Our findings of modest and selective cognitive deficits in children with SDB are consistent with results of previous research showing children with SDB to have poorer performances on a restricted range of neurocognitive assessments administered as part of larger test batteries.^6,9,10,19,20 In several studies, deficits were greatest for language, spatial skills, attention, and organizational skills.^10,21 In our study, the larger SDB group differences found for the PPVT-R than the K-ABC scores may reflect the relative sensitivity of the former test to accumulated verbal knowledge. This test also requires inhibition of impulsive responses—behaviors that are impaired in children with SDB.^16-18,21,22 Impairment of scores on the K-ABC riddles subscale, which requires the child to solve a verbal puzzle, is also consistent with limitations in verbal comprehension and the ability to be selective and to plan in responding to tasks. These deficits are similar to those suggested by weaknesses on the PPVT-R. The K-ABC triangles subscale scores were also impaired in children with SDB. The triangles subscale is the only timed test, and it requires children to plan and organize their responses in solving spatial puzzles, placing demands on multiple executive functions. These deficits are consistent with data from adult studies that show preferential deficits in planning and organizational skills and in efficiency of performance with SDB,^1-3 and with evidence that adults with SDB have a relatively poorer performance on timed tests, including those that make demands on spatial abilities.²³

The pattern of deficits observed suggests that children with SDB are vulnerable to weaknesses in attending selectively to tasks and planning their responses, and these weaknesses contribute over time to more pronounced deficits in accumulated knowledge and academic skills. The latter deficits are illustrated by the SDB group's reduced scores in verbal comprehension and the reductions in academic achievement observed among snorers. Although other investigators have shown deficits in learning, memory, and general intelligence,^5,9,10 these latter studies included children referred for evaluation with high degrees of medical or psychological comorbidity⁵ or did not comprehensively address confounders such as SES and prematurity.^5,9 Data from a subsample of children recruited from the Infant Practices Study¹⁰ showed reductions in both memory and general intelligence in 5-year-old children with SDB symptoms. However, the average scores for the SDB and non-SDB groups were both higher than normative data. These findings were based on parent-reported symptoms, without evidence that deficits were specific to the small number of children with breathing pauses.¹⁰

Although we posited that impairments in sustained attention may underlie deficits in other cognitive functions, we did not find evidence of these deficits when a standardized continuous performance task was used. This agrees with the findings of Beebe et al²¹ and Gottlieb et al,¹⁰ who also did not find significant attentional deficits in SDB when similar measures of visual attention were used. The CPT, originally designed as a tool for evaluating attention-deficit/hyperactivity disorder, may not be sufficiently sensitive in children with mild SDB, who are not overtly somnolent. Alternatively, computer-based vigilance tasks may be less discriminating in current cohorts of children who are now regularly exposed to computer-based video games compared with the older cohorts in whom such tests were originally developed.

The pathophysiological basis for neurocognitive deficits occurring with SDB is thought to relate to exposure to intermittent hypoxemia and sleep disruption. Animal experiments have shown that intermittent hypoxemia produces apoptosis in the cerebral cortex and hippocampus,²⁴ which correlates with spatial learning deficits.²⁵ In humans, such exposures may disproportionately affect planning and organizational abilities (ie, executive functions), which are particularly dependent on the basal ganglia and frontal lobes, areas sensitive to intermittent hypoxemia. Accordingly, in several studies of adults, hypoxemia most closely predicted deficits in executive functions and psychomotor skills, whereas sleepiness predicted attentional deficits.^1,26 One study showed no threshold effect for hypoxemia or the frequency of breathing pauses, suggesting that even mild SDB may have an adverse impact on neurocognitive function.³

We did not observe neurocognitive deficits to be linearly associated with frequency of breathing pauses or severity of overnight desaturation. This differs from the findings from a German study of children that showed that overnight oxygen nadir was associated with selective impairment in arithmetic scores.⁷ The clinical implications of this latter finding are unclear, because no other desaturation variable was associated with impairment in any academic measure, and because oxygen nadir may not provide a reliable estimate of overnight hypoxemia. In contrast, we found that reported snoring frequency was significantly correlated with cognitive and achievement scores. Even in fully adjusted analyses, significant associations were demonstrated between snoring and scores on the K-ABC achievement scale, 6 of the K-ABC subscales, and the PPVT-R. These findings are consistent with literature that has reported cognitive deficits in primary snorers²⁷ and in children with symptoms of SDB who have few breathing pauses.¹⁰ The relatively small number of children with severe SDB might have limited our ability to detect dose-response relationships with indices of SDB severity. Nonetheless, these data suggest that snoring, especially assessed in general samples without high degrees of SDB severity, may provide a useful measure of chronic sleep disruption. Reported snoring at 8 to 11 years of age also may identify children who had experienced more severe SDB at earlier ages, including times of developmental susceptibility to adverse cognitive effects of hypoxemia. Although we were surprised that habitual snoring was not more strongly associated with cognitive deficits than any snoring in the preceding month, it is possible that parents of middle school children underestimate their children's snoring frequency. Regardless of the physiological basis for this observation, our data suggest that obtaining a snoring history may be useful for screening among children at risk for cognitive and achievement deficits.

A previous report from our group⁴ observed a 4-fold higher prevalence of SDB in preterm children, and our present study was designed to test whether this vulnerable group experiences disproportionate morbidity associated with SDB. Preterm children have a range of learning and cognitive deficits occurring in direct proportion to their immaturity,^28,29 including deficits in language abilities, verbal learning, and spatial and organizational skills.²⁸ Previous research in preterm children has not considered whether unrecognized SDB may account for such deficits. Conversely, previous studies of SDB have not considered prematurity as a contributor to SDB-associated comorbidity. Our analyses identify selective cognitive deficits attributable to SDB, even after adjusting for preterm status and birth weight. Furthermore, among children with predominantly mild SDB, SDB-associated comorbidity appears more pronounced in preterm than in full-term children. Although prospective studies are required, these results suggest that children with more limited cognitive reserve and/or who have had developmental or perinatal stresses may be especially vulnerable to mild SDB in childhood. These children may be less able to compensate for SDB-related cognitive deficits, possibly because of diminished plasticity of function after early brain insults.^30-32 Alternatively, a noncausal association between SDB and cognitive dysfunction in preterm children is plausible; ie, developmental and early life stresses could increase a predisposition to SDB by adversely influencing upper airway neuromuscular control and/or airway anatomy. In either case, because some research suggests beneficial effects on cognition with treatment of childhood SDB,^15,33 there may be utility in efforts to more aggressively screen for snoring and SDB in premature children.

The strengths of our study include the large sample size with a wide range of SES and ethnicity and inclusion of sufficient numbers of preterm children to evaluate associations with birth weight and prematurity. The study used objective measures of neurocognitive function and nocturnal cardiorespiratory monitoring and contained robust measures of SES. Study weaknesses include the relatively low proportion of children with severe SDB, preventing a full evaluation of the extent to which marked hypoxemia or sleep disruption may contribute to cognitive deficits. Although validated measures for the OAHI were used that correlate excellently with full in-laboratory measures,⁴ the home sleep studies did not include direct measures of sleep stages or arousal, preventing evaluation of these factors.

Our analyses suggest that some of the reported cognitive and achievement deficits attributable to mild SDB in the literature may have been overestimated owing to confounding by SES and prematurity. However, our data on a large birth cohort are consistent with an association between mild SDB, including snoring, and selective deficits in verbal comprehension and executive functions, with effects relatively greater in preterm children, who may be particularly vulnerable to stresses associated with SDB. Furthermore, our data suggest the potential utility of snoring history as a screening tool, particularly in preterm children, who appear to be at increased risk for both SDB and SDB-associated cognitive deficits.

Correspondence: Susan Redline, MD, MPH, 11100 Euclid Ave, Department of Pediatrics, Case Western Reserve University School of Medicine, Cleveland, OH 44106-6003 (sxr15@case.edu).

Accepted for Publication: September 1, 2005.

Funding/Support: This study was supported by grants HL HL60957, K23 HL04426, M01 RR00080, RO1 NR02707, and RO1070916 from the National Institutes of Health, Bethesda, Md.