DISCUSSION
Our goal was to identify the components of the SRBD survey useful for OSA screening in children with asthma, a patient population with heightened susceptibility to OSA. Our cohort of consecutively recruited children with asthma demonstrated a high prevalence of OSA, confirming results from previous studies. We report four novel findings. First, of the 22 SRBD survey questions, only loud snoring, morning dry mouth and being overweight were associated with OSA. Second, a combined model of reported loud snoring, morning dry mouth and being overweight generated a probability index (MOP) with greater OSA screening accuracy compared to the composite SRBD score. Third, the BMI z-score also had greater OSA screening accuracy compared to the composite SRBD score, but had similar OSA screening accuracy compared with the combined model of reported loud snoring, morning dry mouth and being overweight. Fourth, children with moderate to severe persistent asthma had twofold higher odds of having OSA compared to those with mild persistent and intermittent asthma. These findings confirm the high burden of OSA amongst children with asthma, particularly those with concomitant obesity, and suggest that a simple yet objective anthropometric measure such as the BMI z-score can be used to effectively screen for OSA in children with asthma, thereby identifying those at risk for adverse health consequences.
The objective of initial OSA screening is to identify the persons with a high likelihood of having OSA in a subsequent sleep study. Another rationale is to recognize persons who might have OSA in the context of their underlying disease. For children with asthma, undetected OSA presents an extra‐ordinary high risk for morbidity. Since in-laboratory polysomnography, the only objective and clinically valid method of OSA assessment in children is expensive and often difficult to obtain, a simple yet accurate low-cost approach is critical to identify children with asthma who need sleep studies in a timely manner.
Several studies investigating sleep disordered breathing in children have reported a higher prevalence of OSA in children with asthma compared to the general population1,19–21. Ramagopal et al reported an OSA prevalence of 81% in children with asthma referred for adenotonsillectomy21. Similarly, Kheirandish-Gozal et al studied children with poorly controlled asthma and showed an OSA prevalence of 63%1. Furthermore, a recent study by Ehsan et al reported an OSA prevalence of 47% in a retrospective patient chart review of children with asthma referred for sleep studies2. As seen in our study, the prevalence of OSA was 40%, confirming the relatively higher OSA prevalence in children with asthma. Our result along with those of previous reports suggest that asthma contributes to OSA pathophysiology, which might explain the increased prevalence. Moreover, in our cohort, children with moderate to severe persistent asthma had twofold higher odds of having OSA compared to those with mild persistent and intermittent asthma (Table 4 ). Interestingly, we also found that boys had three-fold higher odds of OSA compared to girls (Table 4 ) possibly due to inclusion of children in the post-pubertal age group22,23 in whom differences in anatomy have begun to emerge resulting in different susceptibilities for OSA. It is likely, however, that the gender difference in predilection for asthma24,25 contributed to the higher odds for OSA in boys compared to girls, but such an inference will need to be examined and verified in studies that have controls without asthma.
Currently, OSA screening in children is often done with the SRBD questionnaire, a 22-item survey derived from the Pediatric Sleep Questionnaire14, for which a score greater than 0.33 indicates a high OSA risk. In the general population, the SRBD questionnaire has a sensitivity of 78% and a specificity of 72% as reported by Chervin et al8. In that study, OSA was defined as an AHI ≥ 1 events/hr. and was enriched with children scheduled to have adenotonsillectomy. Ehsan et al conducted a validation study of the SRBD in children with asthma, which revealed a screening sensitivity of 81.6% and specificity of 14.4% at an OSA definition of AHI≥2 events/hr2. In our study, we found a screening sensitivity of 47.3% and specificity of 68.8% at the same OSA and SRBD thresholds. The differences in our results may be because their study population comprised patients specifically referred for clinical sleep studies unlike our study that investigated consecutive research participants, thus lowering but not eliminating the potential bias related to sleep study referrals. Interestingly, we found that our MOP model that comprised loud snoring, morning dry mouth and BMI z-score was comparatively better than the SRBD. In fact, the BMI z-score alone had an NPV and PPV of 72.2% and 76.3% compared to 65.5% and 51.0% for the SRBD (Figure 3 ). From a clinical utility and practical standpoint, the BMI z-score may provide the best combination of OSA screening accuracy and ease of application. At the optimal cut-off of 2.07, the predictive values suggest that 76% of patients above this threshold will have an AHI≥ 2 events/hr, hence a 24% false positive rate. Conversely, 72% of the patients with BMI z-scores below 2.07 will have an AHI< 2 events/hr, leaving a 28% chance for false negatives. Based on this paradigm, therefore, children with asthma that have BMI z-scores < 2.07, should still be further evaluated for indicators or risk factors for upper airway obstruction including enlarged tonsils, rhinitis, and asthma severity, particularly if they also report loud snoring or have dry mouth in the morning. Bearing in mind that in our cohort, the odds of having OSA was higher in boys compared to girls (Table 4 ), future work will need to explore whether the OSA screening accuracy of the BMI z-score, MOP model and SRBD survey, differs between boys and girls.
Previous studies have shown that children with asthma who have OSA had higher BMI z-scores compared to those without OSA26,27. Our results confirm these reports, as we showed that children with asthma with BMI z-scores ≥ 2.07 had 8-fold higher odds of having OSA compared to those with BMI z-scores < 2.07. Obesity may also contribute to asthma morbidity28–30, although the causal mechanisms remain unclear. Since OSA has been associated with asthma morbidity, obesity related increase in OSA risk may be one of the links between obesity and worse asthma control. On the other hand, increased upper airway inflammation in asthma may lead to higher susceptibility for OSA thus setting up a vicious cycle of cumulative morbidity5,31. While the potential bi-directional OSA-asthma relationship is complex, the shared risk factor in obesity presents a target for mitigating both the OSA and asthma burden32, which as suggested by our study, could be achieved by monitoring BMI z-scores and identifying those at risk for OSA early.