Objective. Hypoxemia, often assessed via pulse oximetry, is associated with neurocognitive deficits in children. The best way to qualify hypoxemia, or which level of hypoxemia already affects cognition, is unknown.
Methods. We assessed the association of pulse oximetry-derived variables that qualify hypoxemia with impaired academic performance in mathematics in a population-based cross-section of 995 primary school children who underwent overnight home recordings of motion-resistant new-generation pulse oximeter saturation (Spo2). Impaired academic performance in mathematics was based on the last school report and defined as grade 4 to 6 on a 6-point scale (ie, approximately the lowest quintile grades).
Results. Of 10 variables under study, only the nadir of the Spo2 values was significantly associated with impaired performance. Categories of this variable representing mild (ie, 91%–93% Spo2; odds ratio: 1.65; 95% confidence interval: 1.06–2.56) and moderate hypoxemia (ie, ≤90% Spo2; odds ratio: 2.28; 95% confidence interval: 1.30–4.01) both were associated with impaired performance in mathematics.
Conclusions. We suggest using the nadir of the Spo2 values in an overnight study to qualify hypoxemia in future studies. This variable may predict neurocognitive deficits in school children. Mild hypoxemia, as yet widely considered benign, may already affect cognition in childhood.
Adverse impacts of chronic or intermittent hypoxemia on development, behavior, cognition, and academic performance have been reported in many well-designed and controlled studies in children with sleep-disordered breathing1–6 or congenital heart disease.7–11 Data for other clinical conditions that result in chronic or intermittent hypoxemia are sparse. In some studies, adverse effects have been noted at even mild levels of oxygen desaturation.1 Application of pulse oximetry to detect chronic or intermittent hypoxemia is a valid, reliable, and frequently used method.12 Several variables that qualify chronic and intermittent hypoxemia can be derived from pulse oximetry and reference ranges used to identify abnormal clinical conditions.13 However, it is yet unclear which pulse oximetry-derived variable best predicts neurocognitive impairments in children and which level of oxygen desaturation is already harmful. The mean and nadir arterial oxygen saturation measured by pulse oximetry (Spo2) have been reported and shown to have some relationship to decreased intelligence,5 impaired attention,5 and hyperactive behavior.1,6
We recently performed a study on the association between sleep-disordered breathing and impaired academic performance in primary school children.14 As part of this study, children were examined for the presence of nocturnal chronic and intermittent hypoxemia using pulse oximetry. Only a weak association between intermittent hypoxemia and impaired academic performance in mathematics was found.14 These results, however, were recently challenged: the inclusion of children with perhaps “abnormal” Spo2 values (91%–95%) in our group defined as normal could have lowered associated risks, thereby underestimating the impact of nocturnal hypoxemia on cognition.15 We thus reevaluated our data, now attempting to form a proper “normoxic” reference and to find the pulse oximetry-derived variable that best predicts impaired academic performance. Because our previous results suggested a relationship between nocturnal hypoxemia and performance in mathematics, we focused on this school subject. To investigate for an independent relationship, we also decided to assess the association between nocturnal hypoxemia and academic performance in a sample of children who were free of congenital heart disease and unlikely to experience sleep-disordered breathing.
The recruitment strategy has been described elsewhere.14,16 In short, 1760 third graders in 27 primary schools located within the city limits of Hannover, Germany, were approached between February 2001 and December 2001, and 1144 children finally participated. The study protocol was approved by the institutional review board, and the regional directorate of education and parents of participating children gave their informed written consent.
An extended version of Gozal’s sleep-disordered breathing questionnaire was used.4,16–18 This questionnaire included questions on gender, age, parental graduation from school (4-point scale: no graduation/primary school, secondary school, high school, or college/university), and frequency of snoring (4-point scale: never, occasionally, frequently, or always). Children, who snored “never” or only “occasionally” were grouped together as nonsnorers; the remainders were categorized as habitual snorers.
All children underwent nocturnal home pulse oximetry. Detailed information on this method has been published elsewhere.13,14 After data acquisition, artifact-free recording periods were identified using data analysis software (Matlab; MathSoft Inc, Cambridge, MA). This software was also used to calculate the median, 10th percentile, and 5th percentile Spo2, as well as the number of desaturations by ≥4% (D4 events) and desaturation clusters (DC events19) within the artifact-free recording period. Recordings with <5 hours of artifact-free recording were excluded. The nadir Spo2 (SATmin) and the number of desaturations to ≤92% (D92 events) and to ≤90% Spo2 (D90 events) were determined manually using information on signal quality, low perfusion, and pulse wave form.13,20 The DC events all were confirmed manually to identify artifacts.
In addition, the average distance from 100% Spo2 and a cumulative hypoxemia score were calculated for each recording. For the cumulative hypoxemia score, the percentage of time within the artifact-free recording period with (1) Spo2 >95%, (2) Spo2 91% to 95%, (3) Spo2 86% to 90%, (4) Spo2 81% to 85%, and (5) Spo2 ≤80% was calculated, then multiplied by 1 (Spo2 >95%), 2 (Spo2 91%–95%), 4 (Spo2 86%–90%), 8 (Spo2 81%–85%), and 16 (Spo2 ≤80), respectively, and added up. Finally, 100 was subtracted from the resulting score so that a recording with the Spo2 always >95% had a cumulative hypoxemia score of 0. These scores were developed to ascertain the area under the Spo2 curve as a way to quantify hypoxemia across individuals in the context of outcomes.
To investigate the accuracy of the data analysis software, we determined the number of D4 events manually as well as automatically in a subsample of 90 children.13 This revealed an excellent correlation (Spearman correlation coefficient: 0.983; P < .001) between manual and automatic analysis. Desaturation indices, defined as events per hour of artifact-free recording, were calculated for D4, D92, D90, and DC events. A customized questionnaire covering various clinical conditions that potentially influence oximetry results was distributed on the day of the recording.13,16 This questionnaire included the following question: “Does your child have any heart disease (eg, cardiac insufficiency)? If yes, please specify.” In 1 school (34 children), test-retest reliability of pulse oximetry-derived variables was tested on 2 consecutive nights.
On the basis of last term’s report form (including grades on a 6-point scale: 1 for “outstanding” and 6 for “failed”), impaired academic performance in mathematics was defined as grade 4 or worse or requirement for special assistance.14
All analyses were done with statistical software (Statistical Package for the Social Science, release 11.0 for Windows; SPSS, Chicago, IL). Descriptive statistics were used to summarize subject characteristics, questionnaire results, and pulse oximetry data. Differences between groups were ascertained with Pearson χ2 test or the Mann-Whitney U test where appropriate. Correlations between pulse oximetry-derived variables obtained during 2 consecutive nights (test-retest reliability) were investigated using Spearman correlation coefficient. To quantify the associations between pulse oximetry-derived variables and impaired academic performance, we calculated odds ratios (ORs) and their 95% confidence intervals using unconditional logistic regression. Logistic models were adjusted for gender (categorical variable), age (continuous variable), maternal and paternal education (2 categorical variables), and class membership (categorical dummy variable). The last factor was found to be significantly related to both pulse oximetry-derived variables and impaired academic performance and thus met criteria for entering this variable as a potential confounder. To control for confounding by congenital heart disease and sleep-disordered breathing, we repeated analyses in nonsnorers and children without a history of heart disease. Associations between habitual snoring, a history of heart disease, and mild as well as moderate hypoxemia were investigated using Pearson χ2 test. P < .05 was considered statistically significant.
A detailed description of the underlying study population and the study sample has been presented elsewhere.14,16 In brief, of 1144 children who returned completed questionnaires, 1115 underwent oximetry recordings. Of these, 995 recordings were considered sufficient for the study purpose. Subject characteristics are given in Table 1. Among children with a history of heart disease, the following diagnoses (numbers in parentheses) were provided by parents: systolic murmur (3), ventricular septal defect (2), pulmonary artery stenosis (1), pulmonary vein stenosis (1), arrhythmia (1), cardiac insufficiency (1), and cardiac valve stenosis (1). Descriptive statistics and correlations for pulse oximetry-derived variables are presented in Table 2.
Nocturnal Hypoxemia and Impaired Academic Performance
As expected, logistic regression analyses showed negative associations (OR: <1) between baseline variables and impaired academic performance in mathematics and positive associations (OR: >1) between desaturation/hypoxemia variables and impaired performance. The baseline SATmin, however, was the only pulse oximetry-derived variable that was statistically significantly associated with impaired academic performance in mathematics (Table 3). There was a negative association between the SATmin and impaired performance showing decreasing odds for impaired performance with increasing SATmin values.
To transform the SATmin into a categorical variable, we used the 75th percentile (94%) to delimit a reference range (94%–100% SATmin) from a mildly (91%–93% SATmin) and a moderately hypoxemic range (≤90% SATmin). The last cutoff value (approximately the 15th percentile) was chosen because it was associated with impaired academic performance in a previous study.14 Logistic regression analysis including this categorical variable revealed a clear dose-effect gradient and a statistically significant association between mild and moderate nocturnal hypoxemia and impaired performance in mathematics (Table 4).
Nocturnal Hypoxemia, Habitual Snoring, and Heart Disease
There were no statistically significant differences between habitual snorers and nonsnorers for any pulse oximetry-derived variable under study (Mann-Whitney U test: all P > .05). In addition, there was no significant difference in the prevalence of mild (habitual snorers vs nonsnorers: 44.4% vs 42.9%) and moderate hypoxemia (19.2% vs 15.3%). Children with a history of heart disease, however, were somewhat more likely to exhibit mild hypoxemia than those without (prevalence of mild and moderate hypoxemia: 80.0% vs 42.7% and 0.0% vs 15.9%; Pearson χ2 test: P = .053). To test for an independent association, we excluded children with a history of heart disease and/or habitual snoring and repeated logistic regression for this subgroup (Table 5). This showed a statistically significant association between moderate but not mild hypoxemia and impaired academic performance in mathematics.
Prompted by a recent review on hypoxemia and neurobehavioral impairments in children in this journal,15 we reevaluated data from a cross-sectional study on hypoxemia during sleep in children.14 Although only moderately stable across 2 consecutive nights, the SATmin was the only pulse oximetry variable that significantly predicted impaired academic performance in mathematics. A clear dose-effect gradient was observed, which concurs with the generally accepted hypothesis of a causal relationship between hypoxemia and neurocognitive impairment.15 In addition to moderate hypoxemia (≤90% SATmin), mild hypoxemia (91%–93% SATmin) was significantly associated with impaired performance. This is surprising given that this level of arterial oxygen saturation is frequent in children.13,21 In addition, the association between SATmin and impaired performance remained significant after excluding children with a history of heart disease or habitual snoring.
Several studies showed associations between neurobehavioral impairment and clinical conditions that likely lead to chronic or intermittent hypoxemia. Only a few studies, however, actually measured oxygen saturation or provided specific information on Spo2 levels. To our knowledge, none has yet examined various pulse oximetry variables for their ability to predict neurocognitive impairment. In 2 studies, the SATmin was used to qualify nocturnal hypoxemia, and associations with impaired intelligence, attention, and behavior were observed.5,6 These and our data suggest that the SATmin is an accurate predictor for neurocognitive impairment in children and may be used as a standard variable in the evaluation of children.
Little is known about mild hypoxemia and its impact on cognition in children. Although associations between neurobehavioral impairments and saturation levels of <90% Spo2 have been reported,1,14 the clinical significance of Spo2 levels between 90% and 95% is unknown. The 2 studies mentioned above observed impaired outcome in children with SATmin values in the lower 90s (90.7% and 90.9%),5,6 and a recent study found associations between mean SATmin values of 94.1% and neurocognitive impairments in children with primary snoring.18 We included children with a SATmin in the upper quartile in a “normoxic” reference group. This approach enabled us to detect a subtle association between mild hypoxemia (SATmin 91%–93%) and impaired academic performance. However, is this finding biologically plausible? Also, why was it only the SATmin that was significantly associated with impaired academic performance, not any of the other pulse oximetry-derived variables? Our study was not designed to investigate mechanisms to address these questions, and comprehensive data from other studies on the impact of mild hypoxemia on cognition are lacking.15 We speculate, however, that a reduced SATmin in a 1-night recording may reflect a “tip of the iceberg” phenomenon, ie, an increased propensity of an individual to desaturate during sleep and occasionally even further than in this 1-night study. Alternatively, mild hypoxemia could be a proxy for unstudied third variables such as sleep disruption/fragmentation, multiple arousal, or nocturnal movement or could lead per se to these pathophysiologic consequences. If this is true, then these factors may have a greater part to play in academic performance than they have been credited for.
Chronic or intermittent hypoxemia is well known to occur in children with congenital heart disease or sleep-disordered breathing and result in neurocognitive impairment. We, however, found no significant differences in the occurrence of mild and moderate hypoxemia in children with and without these conditions. Moreover, there was a strong and significant association between moderate hypoxemia (SATmin ≤90%) and impaired performance in a sample of children who were unlikely to experience congenital heart disease or sleep-disordered breathing. Thus, other factors that result in hypoxemia and neurocognitive impairment likely contributed to this finding. For example, asthma is known to be associated with both hypoxemia22,23 and neurobehavioral impairment.24,25 More studies are needed to investigate potential causes of mild and moderate hypoxemia in otherwise healthy children.
Some of the limitations of the present study have been discussed elsewhere.14 In brief, because of the cross-sectional study design, we were unable to prove a temporal sequence between nocturnal hypoxemia and impaired academic performance as 1 important aspect of causality. We cannot fully rule out that our findings might be affected by sampling bias, unreliable data collection, or incomplete adjustment for confounding factors. In addition, it cannot be fully excluded that the associations reported are chance findings. Multiple explorative significance tests involving several pulse oximetry variables were performed. Notwithstanding this concern, the association between the SATmin variable and impaired performance remained significant after a Bonferroni correction, making the occurrence of a type 1 error unlikely. Sampling bias would require the overrepresentation of children with impaired academic performance exhibiting low SATmin values. As parents were not aware of the exposure (chronic or intermittent hypoxemia), sampling bias was unlikely to occur. This conclusion is also supported by our comparisons for representativeness.16
Using motion-resistant new-generation pulse oximetry, information on signal quality and low perfusion, software-supported manual analysis, and rigorous event definitions including the pulse wave form signal, we accurately identified nighttime chronic or intermittent hypoxemia. An artifact-free recording time of >5 hours was considered sufficient for the study purpose. We, hence, might have failed to detect some hypoxic episodes during the night. This would have led to misclassification and lowered associated risks. We also measured nocturnal saturation across a time span of almost 1 year. Thus, influences related to seasonal variations (eg, colds, allergies) may have occurred. We carefully examined this potential bias but did not find a statistically significant trend (data not shown). Given that the highest prevalence of moderate hypoxemia was found in April (26.9%), one could suspect, however, a relationship between allergies and nocturnal hypoxemia. This should be an objective of future studies.
For the assessment of neurocognitive ability, we used school grades. Grades provide only a rudimentary assessment of cognitive and learning capabilities. The use of teacher ratings, however, was justified by past studies that found relationships between biological risks and similar teacher ratings.4,17 No child was clinically evaluated for the presence of congenital heart disease; neither was sleep-disordered breathing ruled out by a standard evaluation. Hence, some children in the subsample who were suggested not to be affected by either condition may have experienced 1 or both. However, it is difficult to explain why parents should conceal a congenital heart disease or how significant sleep-disordered breathing occurs without its major symptom (ie, habitual snoring). Last, we could account for several social and school-related confounding factors (parental education and teachers’ ratings). Despite this, it is possible that an unstudied variable related to impaired performance was responsible for part or all of the associations reported herein. Sleep disruption, for instance, might be the underlying cause for impaired academic performance, and nocturnal hypoxemia was only a proxy.
This study supports previous reports suggesting that mild and moderate hypoxemia lower cognitive abilities in children. The SATmin was identified as a significant predictor for impaired academic performance and may be used to quantify hypoxemia in future studies. Mild hypoxemia seems to be frequent in primary school children, and more studies are needed to understand its clinical significance. Neurocognitive consequences should be considered if children are exposed to or not withdrawn from chronic or intermittent hypoxemia. Apart from congenital heart disease and sleep-disordered breathing, factors that result in chronic or intermittent hypoxemia should be identified to lower the incidence of mild cognitive impairment in children.
Ms Urschitz-Duprat was supported by a research grant from the Hans Meineke Foundation, Hannover, Germany.
We thank Dr Ehrhardt (Department of Public Health, City Council, Hannover, Germany), Mrs Martinsen (Supervisory School Authority, Hannover, Germany), Mr Hegemann (District Government, Hannover, Germany), and the headmasters and teachers of the participating schools for support and cooperation. Our thanks also go to Robert Downes (getemed AG; Teltow, Germany) and Volker von Einem (Department of Biomedical Engineering, Hannover Medical School, Hannover, Germany) for technical assistance. We particularly thank all of the children and their parents for patience and cooperation; they made this study possible. Oximeter sensors were provided by Masimo Corp (Irvine, CA).
- Accepted October 14, 2004.
- Reprint requests to (C.F.P.) Department of Neonatology, University Hospital of Tuebingen, Calwerstrasse 7, 72076 Tuebingen, Germany. E-mail:
Conflict of interest: Dr Urschitz received a travel grant in 2001 and a research support grant in 2004 from Masimo Corp.
Dr Schlaud’s current affiliation is the Department of Epidemiology and Health Reporting, Robert Koch Institute, Berlin, Germany.
- ↵Gozal D. Sleep-disordered breathing and school performance in children. Pediatrics.1998;102 :616– 620
- ↵O’Brien LM, Holbrook CR, Mervis CB, et al. Sleep and neurobehavioral characteristics of 5- to 7-year-old children with parentally reported symptoms of attention-deficit/hyperactivity disorder. Pediatrics.2003;111 :554– 563
- ↵Bass JL, Corwin M, Gozal D, et al. The effect of chronic or intermittent hypoxia on cognition in childhood: a review of the evidence. Pediatrics.2004;114 :805– 816
- ↵Schlaud M, Urschitz MS, Urschitz-Duprat PM, Poets CF. The German study on sleep disordered breathing in primary school children: epidemiological approach, representativeness of study sample, and preliminary screening results. Paediatr Perinat Epidemiol.2004;18 :431– 440
- ↵Gozal D, Pope DW Jr. Snoring during early childhood and academic performance at ages thirteen to fourteen years. Pediatrics.2001;107 :1394– 1399
- ↵O’Brien LM, Mervis CB, Holbrook CR, et al. Neurobehavioral implications of habitual snoring in children. Pediatrics.2004;114 :44– 49
- ↵Brouillette RT, Morielli A, Leimanis A, Waters KA, Luciano R, Ducharme FM. Nocturnal pulse oximetry as an abbreviated testing modality for pediatric obstructive sleep apnea. Pediatrics.2000;105 :405– 412
- ↵Smith TF, Hudgel DW. Arterial oxygen desaturation during sleep in children with asthma and its relation to airway obstruction and ventilatory drive. Pediatrics.1980;66 :746– 751
- ↵Bender BG, Belleau L, Fukuhara JT, Mrazek DA, Strunk RC. Psychomotor adaptation in children with severe chronic asthma. Pediatrics.1987;79 :723– 727
- Copyright © 2005 by the American Academy of Pediatrics