Introduction

Obstructive sleep apnea (OSA) refers to the occurrence of repetitive episodes of complete or partial upper airway obstruction during sleep, and affects 1–3% of children in the US [1].OSA and primary snoring (PS) can be considered on a continuum, from infrequent partial sleep-related upper airway obstruction without significant hypoxemia or hypercarbia (PS) to apnea with hypoxemia and carbon dioxide retention [2]. Sleep disturbances as a consequence of nocturnal asthma have been reported in children [3], but there is only a small body of literature describing polysomnographic (PSG) features in asthma. Respiratory illnesses such as asthma, which affects 7.5–10% of American children, and atopy are reported to be risk factors for the development of OSA [4–6].

While OSA and asthma are considered separate diagnoses, some have suggested that they may in fact be associated. The pathophysiology of these disorders likely overlap, given that both are affected by inflammation, neural input, and anatomic factors such as obesity [5]. Nocturnal asthma may cause sleep disruption and/or poor quality sleep; unstable asthma treated with oral corticosteroids may increase the risk of OSA [7]. Snoring and noisy breathing are the hallmarks of OSA in children and are also recognized as common complaints of children with asthma [8]. These symptoms are common reasons for referral for PSG evaluation.

The primary aim of this study is to determine whether there is an association between OSA and asthma in pediatric patients referred for PSG for nocturnal snoring. We hypothesized that asthma, as reported on a medical history questionnaire obtained on the night of study would increase the likelihood of OSA. In addition, we expected to find differences in sleep architecture and physiological parameters between asthmatic and non-asthmatic snorers.

Materials and methods

Study sample

Data were obtained from a 1-year (July 2003 to June 2004) retrospective review of clinical charts and overnight PSG results of patients referred to the pediatric sleep laboratory at the University of Maryland for evaluation of obstructive sleep apnea with a chief complaint of snoring. The Institutional Review Board at the University of Maryland approved this study and granted an “exempt protocol” status for the chart review.

Charts and PSG results on 236 patients between the ages of 2 and 15 years were reviewed. Current symptom and history data were obtained from a questionnaire that was completed by the adult caregiver accompanying the child on the night of the study. The questionnaire included patient demographics, the patient’s past medical history (including asthma), medication use, sleep, and daytime behaviors. Asthma case definition was determined by an answer of “yes” asthma (item 1), as a listed diagnosis (item 2), or asthma related medication use (item 3, General Medical Information—Appendix) as previously used in Baltimore children [9]. Medical insurance information was available from hospital intake data. We did not directly collect data on household income or other indices of socioeconomic status. As reported by other authors, we used median household incomes from the zip code of the patients’ residence using US government census information for the year 2000 as an indirect estimation of socioeconomic status [10, 11]. In addition, we coded zip codes falling within the Baltimore Metropolitan Statistical Area as urban.

Only children aged 2 to 15 years, referred to the sleep laboratory for diagnostic studies for the chief complaint of snoring, were included in this study. Even though our laboratory studies patients up to 21 years of age, we chose the age group of 2–15 for analysis to maintain consistency in the application of pediatric criteria for the diagnosis of OSA. Those referred for oxygen, continuous or bilevel positive airway pressure titration, or evaluation of parasomnias were excluded from the analysis. The pediatric sleep laboratory at the University of Maryland serves largely an inner-city population.

Standard overnight monitoring was accomplished with the Alice 4 diagnostic system (Respironics, Murrysville, PA, USA). The patient slept in a quiet darkened room up to 7 h with a parent or guardian present. The montage included EEG leads O1A2, O2A1, C1A2, C2A1, LOC, ROC, submentalis EMG, EKG, measurements of airflow (oronasal thermistor) and end tidal CO2 (ETCO2), chest and abdominal impedance using strain gauges, and oxyhemoglobin saturation using a Nellcor pulse oximeter. Respiratory events were described as: 1) obstructive apneas, with cessations of airflow with continued thoracic and/or abdominal respiratory effort lasting two respiratory cycles in duration; or 2) hypopneas, with reductions of airflow >50% associated with ≥4% fall in oxygen saturation or post-event arousal [12]. The severity of OSA was defined by the apnea–hypopnea index (AHI; number of obstructive apneas and hypopneas per hour of sleep). The arousal index included respiratory and non-respiratory arousals per hour of sleep. Baseline and nadir oxyhemoglobin saturation, baseline and peak ETCO2 in addition to sleep stage were recorded. No patient was acutely ill on the night of the study. Polysomnographic scoring was performed in 30-s epochs according to the Rechtshaffen–Kales criteria [13] by the same technician and reviewed by one of two pulmonologists.

When the height and weight were obtained at the time of the overnight PSG, we were able to calculate the body mass index (BMI) based on the Center for Disease Control nutrition and activity website and is expressed in kg/m2 [14]. BMI percentile was calculated to represent 5th–95th percentile as “normal” and >95th as “overweight” [15]. OSA is defined as an obstructive AHI of five or more events per hour and/or an obstructive apnea index of one or more events per hour [16, 17]. For the purposes of this paper, OSA was defined as AHI > 2.

Analysis

Data were analyzed using Stata/SE for Windows (StataCorp. 2005. Stata Statistical Software: Release 9.2. College Station, TX, USA: StataCorp LP). Data were compiled and where appropriate, expressed as mean ± SD when data were normally distributed (as identified using the Stata command “sktest” [18], group differences between the means were assessed using the two sample t test. However, when data were non-normally distributed, differences between group medians were compared using the Mann–Whitney U test, and are so indicated. In addition, some data were prepared using dichotomous valuations, in which cases chi-square testing was used. Logistic regression was performed to ascertain predictors of OSA controlling for known risk factors such as age, gender, BMI, ethnicity, medical insurance status, and median income for zip code. Differences in OSA prevalences were compared using chi-square. We are unaware of published consensus guidelines for OSA severity in children; therefore, the AHI severity levels used were our best clinical judgment. For some analyses, where indicated, we analyzed data using different definitions of “positive” for OSA: an AHI of >2, or >5 as there is no agreement among clinicians which threshold to use.

Due to the preliminary and clinical nature of the study, in which a type II or beta error is less troublesome and can generate further hypotheses and research questions, we relaxed our assumptions of significance to reject the null hypothesis at the 10% level. In future investigations on this population, the knowledge generated from this preliminary study will permit us to recruit larger samples, therefore permitting us to use more rigorous assumptions of significance.

Results

Clinical characteristics

The average age of the cohort was 7.18 ± 3.74 years (mean ± SD), with most of the children <11 years old (n = 190 or 80.5%). There were more boys (n = 137 or 58%) than girls. Ethnicity data was available on 219 subjects, with the majority being African-American (n = 173 or 79%), followed by Caucasians (n = 39 or 17.8%). Hispanics and Asians were minimally represented (2.3% and 0.9%, respectively). This reflects the socio-demographic profile of Baltimore City in which 20% of the population is between 5 and 17 years and 90% of these children are African-American [19]. The median household income estimated from zip codes and census data was $36, 661 ± $15,669. Most of the children in this study were on medical assistance (n = 171, 72.5%), showing that they came from households in the lower-income brackets.

Because height was not consistently measured on all subjects on the night of the PSG, BMI was calculated on 124 subjects only. The mean BMI was 22.4 ± 8.27 kg/m2, and the mean BMI percentile was 74.3 ± 31.7. Most of the children (71/125 or 56.8%) had BMI calculated at <95th percentile. The remaining 54 children were overweight. Weight was available on 207 subjects; the mean was 79.6 ± 55.2 kg.

Asthma was reported by 74 (31.4%) of the caregivers. Table 1 shows the comparison of asthmatic and non-asthmatic subjects with respect to gender, race, AHI, BMI percentile, median household income by zip code, and insurance status. There were no significant differences in any demographic variable between asthmatics and non-asthmatics. Of the 41 subjects with asthma on whom we could calculate BMI, 20 were obese.

Table 1 Clinical characteristics of asthmatics and non-asthmatics
Full size table

Whether OSA was defined as an AHI >2 or >5 there were no significant differences in OSA prevalence for asthmatics compared to non-asthmatics in bivariate analyses. PSG findings with an AHI < 2 was diagnosed in 139/236 (58.9%) of all subjects (Table 2). The levels of severity that were analyzed were AHI <2, 2 ≤ 5, 5–10, and >10.) OSA, described as AHI ≥ 2 was diagnosed in 97/236 (41.1%) of all snorers referred for a PSG. AHI 2–4.9 was identified in 36 (15.3%), and AHI > 5 was seen in 61 subjects. While we hypothesized that asthma would increase the severity of OSA, the distribution of OSA severity was not significantly different between asthmatics and non-asthmatics in the bivariate analyses (data not shown, p = 0.622). Please refer to Table 3 for a listing and description of all variables used for analysis.

Table 2 Distribution of OSA severity and asthma
Full size table
Table 3 Variable identification and range
Full size table

However, in logistic regressions that included several PSG results as further independent variables and controlled for individual factors and socioeconomic factors (Table 4), we found that reported asthma significantly decreased the odds of OSA (OR [odds ratio] 0.399, 0.160–0.994, p = 0.040).

Table 4 Predictors of OSA
Full size table

In bivariate analysis, African-American children were significantly more likely to have OSA (chi-square 13.05, p < 0.0001), but were no more likely to have asthma (chi-square 0.097, p = 0.76). Because this sample is primarily African-American, it was possible that this finding is artifactual (data not shown). We explored the relationship between race and OSA further in the aforementioned logistic regression. In this regression, we found that African-American children have 2.85 times higher odds of OSA. (Table 4, p = 0.056).. Likewise, children on medical assistance were at increased odds of OSA compared to those with private insurance in the bivariate analysis (chi-square 4.68, p = 0.031), but that finding was not supported in the logistic regression. The AHI was lower in the African-American group, but the difference was not significant 3.3 vs.5.7 obstructive events/hour, p = 0.14). In bivariate analysis, African-Americans were significantly more likely to have an AHI > 5 (chi-square 5.40, p = 0.02) or to have an AHI > 2 (chi-square 20.06, p < 0.0001).

Sleep disruption can be a manifestation of nocturnal asthma. We investigated parameters of sleep fragmentation for children with and without asthma by polysomnography. Total sleep time was significantly decreased in asthmatics at 333 ± 64 min vs. 347 ± 65 min for non-asthmatics (Mann–Whitney U, z = 1.656, p = 0.10). Asthmatics had an increased arousal index (11.3 ± 7.4) compared to non-asthmatics (9.3 ± 6.4, p = 0.099), with no difference in slow wave sleep as a percentage of total sleep time (SWS/TST%), rapid eye movement as a percentage of total sleep time (REM/TST%), sleep efficiency (time asleep as a percentage of total sleep time), or physiological measurements such as baseline and nadir oxyhemoglobin saturation (SaO2%), exhaled carbon dioxide (ETCO2), or time spent with oxyhemoglobin saturations <90% (T90).

Discussion

The main finding in this study is that a history of asthma reported by a snoring child’s guardian decreases the likelihood of OSA, but is associated with mild sleep disturbance. This finding actually was opposite to what we originally postulated. Surprisingly, there was no major difference between asthmatics and non-asthmatics in parameters that would suggest poor sleep. In the ensuing discussion, we review these findings in light of the currently available literature.

Snoring and noisy breathing are recognized as common symptoms in pediatric patients with asthma [8]. In this regard, several theories have been proposed to explain why upper airway abnormalities might constitute a risk for asthma or vice versa. OSA and asthma share risk factors such as allergic rhinitis, gastroesophageal reflux, and obesity [20, 21]. In addition, similar cytokine, chemokine, and histologic changes are seen in both OSA and asthma [22]. Abnormalities of the nasopharynx and lower airways may co-exist due to similar airway responses to inflammatory or atopic stimuli [20]. Sleep deprivation, chronic upper airway edema, and inflammation associated with OSA may further exacerbate nocturnal asthma symptoms [22, 23]. The “one airway” hypothesis suggests that upper airway inflammation and intermittent hypoxia from obstruction may influence the expression and severity of disease of the lower respiratory tract [24]. CPAP treatment of OSA improves asthma control in adults, particularly nocturnal attacks [25] and bronchial hyperresponsiveness [26]. Tonsillectomy and/or adenoidectomy (T&A), rather than CPAP, is the first line of treatment in pediatric OSA and is curative in 75–100% of the cases [5]. It is not known if treatment of OSA changes the course of asthma in pediatric populations.

In spite of the above rationale, self-reported asthma was actually inversely correlated with the likelihood of OSA although not with severity. We speculate that this surprising finding is due to a referral bias. It is likely that parents of a child diagnosed with asthma are more vigilant regarding nighttime symptoms and/or are referred more often due to increased interactions with medical professionals. Since asthma symptoms are often worse at night, noisy breathing may have been mistaken for snoring, prompting a referral for a sleep study. In other words, self-reported asthma predicted a lower likelihood of having OSA. We do not have a biologically plausible explanation for this. We speculate that our unexpected finding is a result of parents and caregivers of asthmatic children being more vigilant to nocturnal symptoms. Since both asthma and OSA can have overlapping nocturnal symptoms, there may be some reporting bias on the part of parents, who might choose to “over” report the presence of asthma in children who snore. We did not perform objective assessments of airways obstruction or bronchial hyperreactivity in this retrospective analysis. Finally, we did not evaluate atopy in this cohort, which could predispose to both asthma and OSA. However, to determine if there is a positive relationship between asthma and OSA, a population-based study would need to be performed.

Thirty-three percent of children in this study had guardian-reported asthma, which is three times the national prevalence rate of 10% [9, 28] and statewide asthma prevalence of 10.6% [29]. This rate is also higher than the 20% asthma prevalence rate in some inner-city Baltimore schools for children from the same catchment area [9]. Data were obtained from a clinical questionnaire that was not validated at the time of this study. However, the majority of the clinical questions were similar to that used in the BASS questionnaire, which is a 50-item validated questionnaire for OSA screening in children [27]. Asthma “yes” as a listed diagnosis on school emergency forms, albuterol or inhaled steroid was a method used in a previous study for case definition of asthma in Baltimore children [9].

Body mass index percentiles of participants in this study revealed that more than 40% were obese with a BMI >95th percentile, which may have influenced the high prevalence of asthma in this snoring cohort. The Cleveland Children’s Sleep and Health Study found that there was a 1.89-fold increased risk of wheeze, but not asthma in 8- to 11-year-old inner city children with sleep disordered breathing (snoring with or without OSA) [15] and that sleep disordered breathing might explain in part the relationship between obesity and asthma.

Asthma and OSA have common risk factors such as African-American ethnicity, obesity, atopy, and poor socio-economic status [8, 20, 30, 31]. Consistent with the literature [5, 6, 21, 32] we found that African-American children were 3.13 times more likely to have OSA and 4.9 times more likely to have asthma than other children. There was also an increased risk of OSA in children on medical assistance compared to those with private insurance. Low SES, severe neighborhood disadvantage, and neighborhood distress are proposed risk factors for childhood OSA [33]. This may reflect increased exposure to environmental triggers of airways inflammation as the housing stock of Baltimore’s inner city is burdened with cockroach and mouse allergen [34], which contribute to symptoms of allergic rhinitis and asthma. We did not find, however, a difference in OSA severity between African-American and Caucasian children referred for PSG. This may be because African Americans were overrepresented in our cohort. While others have reported that asthma increases the prevalence of snoring [8, 31], it has not been previously reported whether or not a history of asthma increases the risk or severity OSA in children. Chi-square analysis showed that a diagnosis of asthma did not increase the severity of AHI (p = 0.549). Indeed, by logistic regression analysis, a self-report of asthma decreased the odds of having OSA (AHI > 2) by 40%, controlling for individual and socioeconomic factors.

There are few data on the polysomnographic findings in subjects with asthma. Initial reports over three decades ago by Kales et al.[35] suggest that children with asthma have a decrease in stage 4 sleep, total sleep time, and frequent awakenings when compared to normal controls. Some authors [36] have demonstrated a greater fall in nocturnal saturation in subjects with stable asthma when compared to healthy individuals [36, 37]. In our study, asthmatic snorers had worse sleep quality as judged by indices of sleep fragmentation than non-asthmatic snorers. This was demonstrated by a slightly higher arousal index, and lower TST. Despite the small increase in arousal index in those with asthma, we are unclear of the clinical relevance of this finding. We do acknowledge the difficulty in scoring of arousals; however, consistency was maintained by having the same technician read all the studies. Increased sleep fragmentation in asthma could also be due to the use of asthma medications such as albuterol, theophylline, and corticosteroids. None of our patients was using these medications the night of the study. While our patients were asymptomatic per parent and technician observation, it is possible that subclinical airways obstruction could have led to greater sleep fragmentation. We did not measure airflow by spirometry during the night to determine if this was the case.

It is unclear if snoring per se alters sleep architecture. In a study of 14 adults, eight “heavy” snorers and six “light” snorers, Hoffstein et al. [38] reported that there was no change in TST, SWS, and REM time between the two groups, but found that snoring intensity did affect sleep efficiency. In another study, the same authors reported that in a group of adult symptomatic snorers and asymptomatic non-snorers, arousals were respiratory in origin in the snorers, but were EEG arousals in non-snorers. The total arousal indexes in both groups, however, were similar [39].

A major limitation of this study was the lack of an objective diagnosis of asthma as well as the fact that there were no data in the questionnaire about the frequency of and severity of the most recent asthma exacerbation. This was a referral-based and not a population-based study, which may explain the high prevalence of reported asthma. A parent report of asthma is the standard for the Behavioral Risk Factor Survey [40], which is a national survey instrument that has been validated. Further study of a cohort of asthmatic children when symptoms are well controlled compared to those with poorly controlled asthma would elucidate further the relationship between pediatric asthma and risk of obstructive sleep apnea.

In conclusion, predominantly African-American inner-city children, referred to the sleep laboratory for an evaluation of snoring, have a high prevalence of self-reported asthma. The odds of OSA were lower in those with prior history of asthma. OSA severity did not change with or without a history of asthma. A higher arousal index in the asymptomatic stable asthmatics suggests a mechanism other than overt nocturnal symptoms of asthma responsible for sleep fragmentation. Objective testing using methacholine or exercise challenges and the use of validated asthma questionnaires to confirm prior vs. current history of asthma will be useful to study the association between asthma and OSA more thoroughly.