Sleep-disordered breathing (SDB) is a group of respiratory disorders observed during sleep, with higher prevalences in preschool and school children[1]. Its main risk factor is adenotonsillar hypertrophy[1,2]. Obstructive sleep apnea syndrome (OSAS) is an SDB subtype. It is defined as a syndrome of upper airway dysfunction during sleep, characterized by snoring and/or increased respiratory effort secondary to increased upper airway resistance[3]. OSAS affects up to 5.8 % of typically developed (TD) children assessed by polysomnography [4–6] and 10.5 % of those evaluated by questionnaires[7]. Most children are approximately 2 to 8 years of age, due to the relative size of the upper airway lymphatic tissue[3]. The most predictive symptoms of OSAS are snoring, respiratory distress during sleep, and parentally witnessed apnea[8]. OSAS confers a risk of developing associated behavioral, learning, metabolic and cardiovascular morbidities[2,9,10,11].

Cerebral palsy (CP) is the most common physical disability in childhood[12]. Children with CP are at increased risk for OSAS [13–15]. In addition to adenotonsillar hypertrophy, children with CP may also present other characteristics that result in airway collapse or obstruction, such as hypotonia of the palate and constrictor muscles, glossoptosis, and laryngeal dystonia [16].

Nocturnal in-laboratory polysomnography is the gold standard test for the diagnosis of OSAS in children with clinical manifestations of SDB [1]. However, in countries where access to polysomnography is poor, assessments from questionnaires are effective clinical tools used to identify patients at increased risk for OSAS [1,17,18]. Some American [5,19,20], European [14,16,21] and Asian [22] studies used questionnaires for assessment and demonstrated higher prevalences of OSAS in children with CP compared to controls.

The Pediatric Obstructive Sleep Apnea Screening Tool (PosaST) questionnaire consists of six questions about pediatric sleep and has high sensitivity and moderate specificity for high risk for moderate to severe OSAS (HR-OSAS) in preschoolers and schoolchildren [1].

This study aimed to: (1) evaluate Brazilian children with CP consecutively treated at the SARAH Network of Rehabilitation Hospitals, regarding the prevalence of HR-OSAS using the PosaST Brazilian version; (2) analyze its association with demographic, clinical and functional variables.

The hypotheses were: (1) the prevalence of HR-OSAS would be significantly higher in children with CP compared to TD children assessed by polysomnography and questionnaire; (2) children with CP with more severe functional impairment would have a significantly higher prevalence of HR-OSAS than those less impaired.

Among the 319 children who met the inclusion criteria, seven were removed after applying the exclusion criteria. The final sample consisted of 312 children (GMFCV I-V). The continuous variable age did not present a normal distribution according to the Shapiro-Wilk test. Most of the 312 children (59.0 %) were aged 6–9 years (median 6.0 years, IQR 5.0–8.0), 164 males, 69.5 % GMFCS I-IV, from the five regions and 14 of the 26 states of the country (Table 1). The GMFCS I-IV sample consisted of 217 children, 60.4 % aged 6–9 years (median 6.0 years, IQR 4.0–8.0), 111 males. The GMFCS V sample comprised 95 children, 55.8 % aged 6–9 years (median 6.0 years, IQR 4.0–8.0), and 54 males. There were no significant differences in age groups and sex frequencies comparing GMFCS I-IV vs GMFCS V samples (Table 2).

The following variables were significantly more frequent in the GMFCS V sample compared to GMFCS I-IV: Friedman criteria 3–4 (OR = 2.75; 95 % CI 0.47, 1.57); hospitalizations for respiratory causes in the last 12 months (OR = 2.45; 95 % CI 0.27, 1.53); more than one hospitalization for respiratory causes in the last 12 months (OR = 3.89; 95 % CI 0.22, 2.51); use of antibiotics in the last 12 months (OR = 1.82; 95 % CI 0.09, 1.11); GERD (OR = 0.26; 95 % CI −2.05, −0.63); drooling (OR = 0.09; 95 % CI −3.03, −1.84); epilepsy (OR = 6.62; 95 % CI 1.35, 2.44); HR-OSAS (OR = 2.51; 95 % CI 0.13, 1.70) (Table 2).

There was a significantly higher prevalence of HR-OSAS in the GMFCS V sample (14.7 %) compared to the prevalence of OSAS in TD children assessed by polysomnography (OR = 0.01; 95 % CI −7.79, −1.86). The comparisons between the prevalence of HR-OSAS in GMFCS I-V and TD children assessed by polysomnography (OR = 1.59; 95 % CI 0.95, 2.67) and PSQ (OR = 0.84; 95 % CI 0.55, 1.28) and between GMFCS V and TD children evaluated by PSQ (OR = 1.48; 95 % CI 0.82, 2.66) showed no significant differences (Table 3).

The multivariate model for HR-OSAS in GMFCS I-V excluding the Friedman criteria as a dependent variable showed significant positive associations with more than one period of antibiotics use for respiratory causes in the last 12 months (OR = 5.54; 95 % CI 0.68, 2.75) and GMFCS V (OR = 3.15; 95 % CI 0.24, 2.05). The multivariate model for HR-OSAS in GMFCS I-V including the Friedman criteria as a dependent variable showed significant positive correlations with Friedman criteria 3–4 (OR = 8.64; 95 % CI 1.18, 3.13) and more than one period of antibiotics use for respiratory causes in the last 12 months (OR = 4.86; 95 % CI 0.45, 2.72). The multivariate model for Friedman criteria 3–4 showed significant positive associations with GMFCS V (OR = 2.51; 95 % CI 0.29, 1.55) and HR-OSAS (OR = 8.24; 95 % CI 1.13, 3.09) (Table 4).

To the best of our knowledge, this is the largest sample of children with CP ever described for OSAS assessment. The final sample was composed of children from the five regions of Brazil and 14 of the 26 states. As CP is a syndrome with a variable spectrum of severity, the sample was analyzed in its entirety (GMFCS I-V) and separately in GMFCS I-IV and GMFCS V samples.

The availability of polysomnography is quite low in the Brazilian public health system, similar to other low and middle-income countries (LMICs), where assessments from questionnaires are effective for identifying patients at increased risk for OSAS[1,17]. A study examined the association between the results of polysomnography and PSQ in 745 Turkish patients aged 2–18 years with multiple diagnoses, the CP group (80 individuals) had the highest PSQ sensitivity (88.8 %) for the diagnosis of OSAS[18]. Some other studies used the PSQ [15,16] and the Sleep Disturbance Scale for Children (SDSC) in children with CP [14,19,20,22,24]. The PosaST was specifically constructed and validated for the screening of risk for OSAS. Although PSQ (22 items), SDSC (25 items), and PosaST (6 items) are validated and adapted for Brazilian Portuguese and culture, and PSQ is the most sensitive screening questionnaire for OSAS in children [25], PosaST was used in the present study because it is shorter, requiring less time and effort from parents to answer questions, making it more suitable for low-resource settings in Brazil. Furthermore, the six behavioral items of the PSQ do not allow us to differentiate whether these symptoms would be due to OSAS or primary comorbidity in children with CP, which may generate an interpretation bias when used in this population.

The main risk factor for SDB in children is adenotonsillar hypertrophy[1,2]. Respiratory illnesses are the main cause of hospital admissions [26] in CP. A consensus statement on respiratory disease in CP defined the following risk factors: GMFCS V; at least one hospital admission for respiratory illness in the past year; at least two courses of antibiotics for chest infections in the past year; dysphagia; current seizures; daily cough, or weekly wheeze, phlegm or gurgly chest; GERD; mealtime respiratory symptoms; snoring every night[27]. Based on these data, the variables to be analyzed in the present study were defined.

The prevalences of HR-OSAS were 9.0 % (GMFCS I-V) and 14.7 % (GMFCS V). In 2008 a study evaluated 998 Brazilian schoolchildren (9–14 years) via a specific questionnaire about symptoms of SDB [28], but it was not possible to use it for comparison with the present data due to the incompatible age ranges (3–9 years vs 9–14 years). Significantly higher risk factors for OSAS were found among children with CP and CP with epilepsy than among children with other developmental disabilities in 215 American children evaluated via PSQ[15]. Regarding SDB (OSAS included, but not specifically), some studies found significantly higher frequencies in children with CP (12 % to 44 %)[14,16,19,20–22,24]. SDB was significantly more prevalent in 94 Turkish children with CP aged 2–18 years in PSQ scores compared to 94 TD pairs[16]. SDB was the most common sleep disorder in 100 Egyptian children with CP evaluated via questionnaires[13]. SDB symptoms were significantly more frequent in 61 Polish children with CP than in the control group studied via questionnaires[21]. Among 173 Irish children with CP, 14.5 % experienced SDB according to SDSC[24]. Forty-one American children with CP had significantly higher SBD-PSQ scores than TD controls aged 8–12 years[19]. One hundred and nine Malaysian children with CP scored significantly higher than their healthy siblings on SDB/SDSC subscale[22]. SDB prevalences of 14.0 % and 25.6 % were found respectively in 165 Italian children with CP aged 6–16 years [14] and in 43 Brazilian children with CP aged 4–18 years [20] using the SDSC.

Studies conducted in the USA [4,5] with respectively 110 and 700 TD children evaluated through polysomnography (using AIH > 5 as the diagnostic criterion) and in China [6] with 619 TD children (using the International Criteria of Sleep Disorders version II criteria) reported OSAS prevalence rates of 1.2 % to 5.8 %. In a study analyzed via PSQ 1642 Italian TD children, 10.5 % were considered at risk for OSAS[7]. The prevalence of HR-OSAS found in the present study in the GMFCS V (14.7 %) sample was significantly higher than that reported by the studies that used polysomnography (1.2 % to 5.8 %) [4–6], but did not show a significant difference comparing with that one in which PSQ was used (10.5 %)[7]. The prevalence of HR-OSAS found in the GMFCS I-V sample (9.0 %) was not significantly higher than those of the studies that used polysomnography (1.2 % to 5.8 %) [4–6] and PSQ (10.5 %)[7].

No significant differences were found in the frequencies of age groups, sex and more than one period of antibiotic use for respiratory causes in the past year when comparing GMFCS I-IV vs GMFCS V. Contrary, Elsayed et al. (2013) found SDB significantly more frequent in school children than in preschoolers with CP[13]. The following variables were significantly more frequent in the GMFCS V than in the GMFCS I-IV sample: Friedman criteria 3–4, hospitalizations, more than one hospitalization and antibiotic use for respiratory causes in the past year, GERD, drooling and epilepsy. Based on the risk factors described above, the authors already expected to find these significant differences between the samples.

Among the confirmed hypotheses was the significantly higher prevalence of HR-OSAS in GMFCS V (14.7 %) compared to GMFCS I-IV (6.4 %). Other studies also found higher prevalences of OSAS with increasing GMFCS levels[13,15].

The multivariate model for HR-OSAS including the Friedman criteria as a dependent variable showed significant positive correlations with Friedman criteria 3–4 and more than one period of antibiotics use for respiratory causes in the last 12 months, but this model did not find a significant association with GMFCS V. Taking this finding into consideration, the authors constructed a new multivariate model for HR-OSAS excluding the Friedman criteria as a dependent variable. This new model showed significant positive associations with more than one period of antibiotics use for respiratory causes in the last 12 months and GMFCS V. Thinking about the differences found between the two models, the authors developed a multivariate model for Friedman criteria 3–4 that showed significant positive correlations with GMFCS V and HR-OSAS. Therefore, the Friedman criteria 3–4 variable had a significant correlation with HR-OSAS, regardless of GMFCS, and the GMFCS V variable became significantly associated with HR-OSAS in the GMFCS I-V sample at the expense of the significantly higher prevalence of Friedman criteria 3–4 in children classified as GMFCS V. Koyuncu E et al. (2017) also did not find significant relationships between SDB, age and sex [16]. Newman et al. (2006) did not find significant correlations between demographic and clinical variables[24].

AIH, apnea-hypopnea index; CI, confidence interval; CP, cerebral palsy; GERD, gastroesophageal reflux disease; GMFCS, Gross Motor Function Classification System; HR-OSAS, high risk for obstructive sleep apnea syndrome; IQR, interquartile range; LMICs, low and middle-income countries; OR, odds ratio; OSAS, obstructive sleep apnea syndrome; PosaST, Pediatric Obstructive Sleep Apnea Screening Tool; PSG, polysomnography; SD, standard deviation; SDSC, Sleep Disturbance Scale for Children; SDB, sleep-disordered breathing; SRDB, Sleep-related Breathing Disorder Scale; TD, typically developing.

The authors.