Sleep Outcomes in Children with Hemifacial Microsomia and Controls: A Follow-Up Study

Abstract

Objective

Children with craniofacial anomalies are at high risk of sleep-disordered breathing (SDB), yet its prevalence among children with craniofacial conditions is not known. Children with a relatively common congenital craniofacial condition, hemifacial microsomia (HFM), are likely particularly vulnerable to SDB due to underdevelopment of the mandible and oropharynx. Nevertheless, most children with HFM are not referred for sleep studies. We hypothesized that sleep outcomes would be worse in children with HFM versus controls.

Methods

We conducted a follow-up study among 124 cases and 349 controls who previously participated in a study of HFM risk factors. Cases were eligible if diagnosed by a craniofacial specialist at one of 25 participating centers. Controls were matched to cases by pediatric practice and age at interview. Parents completed the Pediatric Sleep Questionnaire (PSQ) regarding symptoms of SDB and sleep habits. Regression models were adjusted for region, age, sex, race/ethnicity and maternal education.

Results

Snoring was more commonly reported for children with HFM (29%) than controls (17%) [adjusted odds ratio (adjOR)=1.9, 95% confidence interval (CI) 1.1–3.2]. Compared with controls, children with HFM more often had symptoms consistent with SDB (adjOR=2.8, CI 1.5–5.1). On average, cases’ parents reported 1.9 times as many symptoms on the PSQ breathing scale (CI 1.4–2.7), and 1.3 times more symptoms on the PSQ sleepiness scale (CI 0.9–1.8) than did controls’ parents, with little difference on the PSQ behavior scale. Parents of children with HFM reported 1.4 times more night awakenings than did controls’ parents (CI 0.9–2.2). There was no association for sleep-onset latency or time in bed.

Conclusions

Children with HFM experienced more snoring and other symptoms of SDB than controls. Pediatricians should be aware of the increased vulnerability for SDB among children with mandibular or external ear underdevelopment or asymmetry, and should refer to a sleep specialist as needed.

Introduction

Children with craniofacial anomalies are at high risk of sleep-disordered breathing (SDB), a continuum of breathing disturbances that includes primary snoring and sleep apnea,¹ yet its prevalence among children with various craniofacial conditions is not known. Children with a relatively common congenital craniofacial condition, hemifacial microsomia (HFM), are likely particularly vulnerable to SDB due to underdevelopment of the mandible and oropharynx.

Occurring in an estimated 2.0 to 3.3 per 10,000 live births,² HFM is characterized by asymmetric underdevelopment of facial structures, including the ear, maxilla and mandible, soft tissues and facial nerves.^3–8 HFM is phenotypically heterogeneous, ranging from microtia with mild mandibular asymmetry to Goldenhar Syndrome, which includes vertebral and eye anomalies^4,8–12 and, variably, extracraniofacial anomalies.^6,8,10–14 The potential sequelae of SDB include learning deficits, behavioral problems,^15–23 and health-related outcomes such as failure to thrive,^24,25 making it important to understand the association of SDB with HFM.

While existing studies have indeed suggested an increased risk of SDB in relation to HFM, these studies are limited by small sample sizes, wide age ranges, lack of control groups, and reliance on medical records to ascertain airway disturbance or sleep outcomes.^26–29 We conducted the first large-scale epidemiologic study comparing sleep outcomes in children with and without HFM. We hypothesized that children with HFM experience a higher prevalence of snoring and SDB compared with controls. Confirmation of this hypothesis would suggest that pediatricians caring for children with mandibular or external ear underdevelopment or asymmetry, which might represent undiagnosed HFM, should have an increased awareness of the potential for SDB in these patients, and refer for evaluation as necessary.

Methods

Study Sample

Eligible subjects were participants in a completed multi-center case-control study of risk factors for HFM, which has been described previously.^14,30 Briefly, cases were enrolled at 26 craniofacial centers throughout the US and Canada from 1996 to 2002, and included children through three years of age with a diagnosis of HFM (including Goldenhar syndrome and isolated microtia or anotia), as determined by their craniofacial specialist and confirmed by medical record review. Cases with known Mendelian inheritance, chromosomal anomalies, or isotretinoin exposure in utero were excluded, as were adoptees. Controls were matched by age at interview (within two months), and were identified through the case’s pediatrician or a similar practice (e.g. based on size and location). Controls were excluded if they had major malformations or had been adopted. Participants in the original study were eligible for follow-up, except at one site due to its Institutional Review Board restrictions (45 cases, 13 controls). We recontacted parents of eligible subjects when the children were between 5.5 and 8.5 years of age. Cases with unconfirmed HFM were excluded from analyses (n=3). This paper is based on participants enrolled between 2005 and 2007, for whom sleep outcomes were collected as a supplement within a larger follow-up study of psychosocial outcomes in children with and without HFM.

Data Collection

Sociodemographic characteristics included child’s year of birth, sex, years of maternal education and family income from the follow-up study, and race/ethnicity from the case-control study. At the time of follow-up, a packet of questionnaires was mailed to parents of case and control children. The packet included questions about medical history; the Pediatric Sleep Questionnaire (PSQ),³¹ which is a 22-item symptom inventory designed to screen for SDB in the general pediatric population; and several instruments to assess psychosocial functioning. Results regarding psychosocial functioning are outside the scope of this report, with analyses ongoing. The follow-up study was approved by the Institutional Review Board at Boston University in Boston, Massachusetts and at each study site.

Determination of HFM severity, presence of associated anomalies, and laterality were based on medical records and photos collected during the case-control study.³² Severity was based on craniofacial specialists’ assessment of the defect (mild and moderate/severe) as noted in medical records. Photographs were used to determine laterality of ear anomalies, including ear and cheek tags, microtia or anotia. We combined these photo ratings with medical records notes to classify HFM as unilateral or bilateral.

Snoring and other symptoms consistent with SDB were assessed by using the PSQ, which referred to symptoms during the previous four weeks.³¹ Snoring was defined as loud snoring or snoring at least half the time. A child was classified as having screened positive for SDB (herein referred to as PSQ+) if at least 8 of the 22 (or 33% non-missing items) symptoms were endorsed. Questionnaires with more than 50% missing items were excluded. The PSQ includes several subscales: breathing, sleepiness, and behavior.

Supplementary Items

We added supplementary questions to the PSQ partway through the study, thus these items were available for a subset of participants. To assess sleep-wake habits, we ascertained number of night awakenings (‘How many times per night does your child usually wake up?’), time in bed (difference between reported bed and wake times), and sleep-onset latency (‘How long does it usually take your child to fall asleep each night?’). Typical latency was defined as 10–30 minutes.

We ascertained aspects of each child’s sleep-related health, including any history of polysomnography, physician-diagnosed SDB or obstructive sleep apnea syndrome, or airway interventions [adenoidectomy or tonsillectomy (AT); continuous positive airway pressure (CPAP); nasopharyngeal airway; tracheostomy]. We also asked for any medication use or respiratory illness in the previous four weeks.

Finally, due to health care for their child’s HFM, case-parents may have an increased awareness of their children’s sleep compared with control-parents. To help assess differential reporting by case status, we therefore asked: (1) ‘On average, how many times per night do you check on your child?’ and (2) ‘How concerned have you been about your child’s sleep?’ (not at all; a little bit; somewhat; quite a bit; very).

Statistical Analyses

We fit unconditional logistic regression models to compare the prevalence of snoring, PSQ+, and typical sleep-onset latency in children with and without HFM. We fit linear regression models to compare average sleep-onset latency and time in bed. We fit negative binomial regression models to compare the number of night awakenings and number of endorsed symptoms for each PSQ sub-scale. The exponentiated coefficients from these models can be interpreted as the ratio of expected awakenings or symptoms in cases versus controls. We computed 95% confidence intervals (CI) for all estimates. For adjusted models, we a priori selected covariates to include in regression models, including sex, age, region (northeast/mid-Atlantic, midwest, south, west), race/ethnicity (Hispanic, non-Hispanic white, other) and maternal education as a marker for socioeconomic status (<12, 12, 13–15, 16+ years).^{1,14,30,33,34}

Because of the reduced sample with supplementary data, the number of covariates was large relative to the number of cases positive for each sleep outcome. This situation can cause multivariate adjusted statistical models to become unstable. To reduce the number of covariates, we adjusted analyses addressing sleep-wake habits by using a propensity score approach. The propensity score was the probability of HFM given sex, age, region, race/ethnicity and maternal education.³⁵ We chose the covariate parameterizations such that the covariate distributions for cases and controls were similar within each quintile of the propensity score. Indicator variables denoting each propensity score quintile were included in regression models in lieu of individual terms for each covariate.

Subanalyses

In order to identify HFM subgroups that may be at particularly high risk for adverse sleep outcomes, we conducted subanalyses to compare the prevalence of snoring and PSQ+ according to severity, presence of associated major anomalies, and laterality. Due to the small number of cases in many subgroups, we present unadjusted estimates for these subanalyses, although adjustment by the propensity score method described above attenuated point estimates slightly and produced similar 95% CIs. In addition, we combined cases with isolated microtia and mild HFM into a single analytic group.

Sensitivity Analyses

To account for recent health history, and to address possible information bias, we conducted a series of sensitivity subanalyses which were possible only among the subsample with supplementary sleep data. Therefore, we first repeated the main analyses in this subsample, followed by stratified analyses according to: recent medication use; recent respiratory illness; and history of AT. To explore whether any differences in sleep outcomes could be driven by children with severe SDB requiring intervention (e.g. CPAP) or be influenced by polysomnography history, we conducted analyses excluding subjects with either. Finally, to account for potential biases due to differential parent awareness of sleep, we compared cases and controls within strata based on similar ratings of sleep-related parent concern. The propensity score method was used in these subanalyses to account for potential confounding by the sociodemographic factors listed above. We used unadjusted exact logistic regression when indicated by small cell sizes.

All analyses were performed using Intercooled Stata 10.0.

Results

We mailed questionnaire packets to 167 and 482 parents of eligible cases and controls, respectively. Among cases, 134 (80%) parents participated, with 7 refusals and 26 non-responders. Among controls, 360 (75%) parents participated, with 30 refusals and 91 non-responders. Eleven case and 37 control families could not be located. Parents of 10 cases and 11 controls completed an abbreviated packet, mainly due to time constraints, and thus were missing the PSQ. The full analytic sample therefore included 124 case and 349 control participants, whereas the subsample who received and returned the modified packet that included supplementary sleep items comprised 84 cases and 248 controls.

Within the full sample, proportionally more cases were male, cases were somewhat older, and cases were more likely to be Hispanic than were controls (Table 1). More case-mothers did not complete high school than control-mothers, and annual household income tended to be lower among case families.

Table 1.

Select sociodemographic characteristics for participants with and without HFM.

	HFM N=124		Control N=349
Characteristic	N	(%)a	N	(%)a
Sex
Female	49	(40)	187	(54)
Male	75	(60)	162	(46)
Year of Birth
1996–1997	30	(24)	59	(17)
1998	41	(33)	115	(33)
1999	28	(23)	100	(29)
2000	25	(20)	75	(21)
Race/Ethnicity
Non-Hispanic White	83	(67)	263	(75)
Hispanic	24	(19)	21	(6)
Other	17	(14)	65	(19)
Region
Northeast/ Mid-Atlantic	30	(24)	93	(27)
Midwest	99	(27)	276	(36)
South	6	(16)	14	(17)
West	10	(33)	22	(21)
Maternal Education b
<12	16	(13)	21	(6)
12	26	(22)	79	(23)
13–15	30	(25)	85	(25)
16+	47	(39)	160	(46)
Annual Household Income c
<$25,000	23	(19)	29	(9)
$25,000 – 34,999	16	(13)	34	(10)
$35,000 – 64,999	24	(20)	85	(25)
≥$65,000	58	(48)	190	(56)

^a% may not add up to 100 due to rounding error

^bMaternal education is missing for 5 (4%) cases and 4 (1%) controls

^cIncome is missing for 3 (2%) cases and 11 (3%) controls

Most cases had moderate/severe HFM (N=73), 33 had mild HFM, and 18 had anotia or microtia. Among the 45 cases with associated anomalies, 36 (80%) had moderate/severe HFM. Of the 97 cases with available medical records notes or photographs, 38 (39%) had right unilateral, 37 (38%) left unilateral, and 22 (23%) had bilateral involvement.

Snoring and Sleep-Disordered Breathing

Snoring was more commonly reported for children with HFM (29%) than controls (17%) (adjOR=1.9, CI 1.1–3.2). Children with HFM were also more often PSQ+ than controls (adjOR=2.8, CI 1.5–5.1) (Table 2). The number of endorsed symptoms on the PSQ was higher on the breathing scale (CI 1.4–2.7), and slightly higher on the sleepiness scale (CI 0.9–1.8), with no appreciable difference on the behavior scale (Table 3).

Table 2.

Crude and adjusted odds ratios comparing the odds of snoring and screening positive on the PSQ in children with and without HFM.

	HFM N (%)	Control N (%)	Crude OR a (95% CI)	Adjusted OR b (95% CI)	p c
Snoring	34 (29)	55 (17)	2.0 (1.2, 3.3) d	1.9 (1.1, 3.2) e	0.021
PSQ+	29 (23)	32 (9)	3.0 (1.7, 5.3)	2.8 (1.5, 5.1) f	0.001

^aOdds Ratio (95% Confidence Interval)

^bAdjusted for maternal education, region, race/ethnicity, sex, and age

^cp-values based on likelihood ratio test statistics

^dMissing for 5 (4%) cases and 16 (5%) controls

^eExcludes 10 (8%) cases and 20 (6%) controls with missing covariate data

^fExcludes 10 (8%) cases and 19 controls (5%) with missing covariate data

Table 3.

Negative binomial regression analysis comparing the number of symptoms endorsed for each subscale of the PSQ in children with and without HFM.

PSQ Subscale	Crude RR a (95% CI)	Adjusted RR b (95% CI)	p c
Breathing	1.9 (1.4, 2.7) d	1.9 (1.4, 2.7)	<0.001
Sleepiness	1.2 (0.9, 1.7)	1.3 (0.9, 1.8)	0.185
Behavior	1.1 (0.8, 1.4)	1.0 (0.8, 1.3)	0.885
Total	1.3 (1.1, 1.6)	1.3 (1.0, 1.6)	0.021

^aRR = expected ratio of events in cases versus controls (95% Confidence Interval)

^bAdjusted for maternal education, region, race/ethnicity, sex, and age; excludes 4 cases and 3 controls with missing covariate data

^cp-values based on likelihood ratio test statistics

^dExcludes 1 control with more than 50% missing items on breathing subscale

Severity, Presence of Associated Anomalies, and Laterality

Compared with controls, cases with moderate/severe HFM more often reported snoring (OR=2.9) and PSQ+ (OR=4.3), which is in contrast to the attenuated odds ratios observed for children with microtia or mild HFM (adjORs=1.0 and 1.6, respectively) (Table 4). When stratified according to the presence of associated anomalies, the observed ORs were larger in cases with versus without associated anomalies, but was similar for unilateral and bilateral cases. Adjustment using the propensity score produced similar results (not shown).

Table 4.

Odds ratios comparing the odds of snoring and of screening positive on the PSQ in children with versus without HFM, according to severity, presence of associated anomalies and laterality.

	Snoring			PSQ+
	Yes N	No N	OR (95% CI)a	Yes N	No N	OR (95% CI)
Controls	55	278	Referent	32	317	Referent
HFM Subgroups
Severity
Microtia/Mild HFM	8	40	1.0 (0.5, 2.3)	7	44	1.6 (0.7, 3.8)
Moderate/Severe HFM	26	45	2.9 (1.7, 5.1)	22	51	4.3 (2.3, 7.9)
Associated Anomalies
Not Present	17	58	1.5 (0.8, 2.7)	16	64	2.5 (1.3, 4.8)
Present	17	27	3.2 (1.6, 6.2)	13	31	4.2 (2.0, 8.7)
Laterality
Unilateral	24	47	2.6 (1.5, 4.6)	19	56	3.4 (1.8, 6.3)
Bilateral	7	14	2.5 (1.0, 6.6)	6	16	3.7 (1.4, 10.2)

^aCrude odds Ratio (95% Confidence Interval)

Supplementary Sleep Questions

Proportionally more parents of children with HFM than controls reported medication use or respiratory illness within the past four weeks, whereas exposure to environmental cigarette smoke was similar in each group (Table 5). Overall, parents of cases (versus controls) more often reported a history of airway interventions, overnight sleep studies (i.e. polysomnography), and diagnoses of SDB or obstructive sleep apnea syndrome. Nearly 20% of children with HFM underwent polysomnography versus 2% of controls. Family history of SDB was fairly similar in cases and controls.

Table 5.

Parent report of sleep-related health history and sleep habits for children with and without HFM.

	HFM N=84		Control N=248
	N	(%)	N	(%)
Recent Health Historya
Medication Use	42	(50)	94	(38)
Respiratory Illness	29	(35)	71	(29)
Exposed to Cigarette Smoke	9	(11)	23	(9)
Past Airway Interventionsb
Tonsillectomy	12	(14)	21	(8)
Adenoidectomy	11	(13)	29	(12)
Nasopharyngeal airway (NP)	2	(2)	1	(0)
Continuous Positive Airway Pressure (CPAP)	6	(7)	5	(2)
Tracheostomy	7	(8)	0	(0)
Sleep Historyc
Polysomnography (PSG)	15	(18)	4	(2)
SDB/OSAS Diagnosis	10		4
Family History of SDB/OSAS	10	(12)	39	(16)
Sleep Habitsd	Mean	(sd)	Mean	(sd)
Time in Bed (hours)	10.3	(0.9)	10.3	(1.1)
Sleep-Onset Latency (minutes)	19	(15)	17	(12)

^aMedication use missing for 1 case and 1 control; respiratory illness missing for 2 cases and 1 control; exposure to cigarette smoke missing for 7 cases and 19 controls

^bTonsillectomy, tracheostomy missing for 1 control; Adenoidectomy, CPAP missing for 1 case and 1 control; NP missing for 2 controls

^cSDB = Sleep-disordered breathing; OSAS = Obstructive Sleep Apnea Syndrome; PSG, SDB diagnosis missing for 3 controls; family history missing for 1 case and 2 controls

^dTime in bed missing for 2 controls; sleep-onset latency missing for 3 cases and 11 controls

Sleep Habits

Mean time in bed and sleep onset latency were comparable in cases and controls (Table 5). Children with HFM were somewhat less likely (70%) than controls (76%) to report a typical time to sleep onset (OR=0.7, CI 0.4–1.3). On average, parents of children with HFM reported 1.6 times more night awakenings than did parents of controls (CI 1.0–2.4).

Parental Concern

Parents of cases checked on their children an average of 1.4 times more often than did parents of controls (CI 0.9–2.2) (not shown). The vast majority of case (70%) and control (81%) parents reported no concern about their child’s sleep, but case-parents were 1.7 times more likely to report at least ‘a little bit’ of concern than were control parents (CI 0.9–3.1).

Sensitivity Analyses

As compared with the full analytic sample, the OR for snoring (2.4, CI 1.3–4.6) was slightly higher in the subsample, while the OR for PSQ+ (2.8, CI 1.3–5.8) was similar (not shown). These estimates did not differ substantively when stratified by recent medication use or respiratory illness. The increased odds of PSQ+ among cases was accentuated among children with past AT (OR=6.1 versus 2.4 in those without AT), with similar results for analyses of snoring (not shown). When excluding children with a history of airway interventions (other than AT) and polysomnography, the increased risk of snoring and PSQ+ in children with HFM versus controls persisted (not shown). Stratification by level of parent concern resulted in somewhat disparate OR estimates for snoring (OR_none=1.8; OR_{≥a little bit}=2.9) and PSQ+ (OR_none=1.6; OR_{≥a little bit}=3.4) (not shown). Similar results were observed within strata defined by frequency of night checks.

Discussion

HFM is a craniofacial anomaly typified by asymmetrical underdevelopment of facial structures, including the mandible. HFM-related abnormalities in the underlying size, position, and morphology of craniofacial bones can secondarily impact soft tissues of the upper airway.²⁷ Any resulting disturbances in breathing often manifest during sleep as snoring or SDB. Despite the plausibility of an association and the use of surgical treatments for sleep apnea among children with HFM,^27,36,37 there have been few studies of airway obstruction or sleep in this group. Therefore, we undertook a study to assess symptoms of SDB and sleep-wake patterns in children with HFM. Our study is the largest known study of sleep outcomes in this population, and one of few such studies to include a comparison group of unaffected children.

Our results are consistent with observations from previous studies in which the prevalence of SDB was relatively high among children with craniofacial anomalies, including HFM.^26–29 The observed increase in nightly arousals is also compatible with reports that snoring and SDB are associated with restless sleep or increased night-time arousals in the general pediatric population.^38–40 Yet, we did not observe differences in sleep patterns such as sleep-onset latency or time in bed between children with and without HFM. Though the observed time spent in bed is consistent with sleep habits described in this age group,⁴¹ misclassification may obscure true differences, if they exist. Alternative methods such as actigraphy and sleep diaries may improve our understanding of sleep patterns among children with HFM.^40,42–44

Symptoms of snoring and SDB were elicited via a questionnaire mailed to participating parents. Although an imperfect indicator of SDB, the PSQ has several strengths. First, it accounts for variation in sleep over time by referring to symptoms during the past four weeks. In addition, unlike the predominantly accepted gold standard, polysomnography, the PSQ is inexpensive, presents minimal participant burden, and can be completed in approximately 5 to 10 minutes. The PSQ also has been validated against polysomnography in a general pediatric clinical population, with estimated sensitivity for predicting SDB of 0.81 and estimated specificity of 0.87.³¹ Compared with the questionnaire, use of polysomnography would likely have led to reduced participation rates in the overall study, and might have introduced bias due to differential participation of children with and without symptoms of SDB.

Nevertheless, there are potential limitations due to parent report, particularly the possibility of biased ascertainment of snoring and SDB status. Specifically, because children with HFM were more likely to be evaluated for sleep-related problems via polysomnography in our sample, case-parents may have an increased knowledge of sleep-related symptoms, which could result in differential reporting by case versus control parents. However, we continued to observe an increased risk of snoring and PSQ+ in sensitivity analyses which addressed this possibility (e.g. those restricted to subjects without a history of polysomnography). Moreover, case-control differences were strongest when expected to be so: the magnitude of the OR was greatest for the PSQ breathing subscale, whereas cases and controls scored similarly on the behavior scale. Furthermore, SDB symptoms were reported to be more severe for cases with more severe HFM, although not according to laterality. Taken together, these sensitivity analyses suggest that bias due to underlying case-parent perceptions or heightened concern is unlikely to explain the entirety of the observed association between HFM and SDB.

We also addressed health-related factors that may have influenced the magnitude of observed associations, such as recent respiratory illness. In virtually all such analyses, point estimates of associations were altered only minimally, suggesting that the study results are robust to various sources of bias. Given the empirical data and the biological plausibility of the association, it is unlikely that the observed association between HFM and SDB is spurious.

We attempted to increase statistical precision by using a propensity score method to limit the number of covariates in regression models. Yet, subanalyses to compare the prevalence of snoring and PSQ+ according to case characteristics were limited by small cell sizes and imprecise point estimates, necessitating larger studies to confirm observed trends.

Even in the absence of rarer, severe sequelae such as failure to thrive, persistent SDB is thought to worsen children’s general health and behavior, possibly inducing sleepiness, hyperactivity, and learning problems.^7,25,45 These consequences may be why the American Thoracic Society has recommended that SDB should be evaluated in at-risk children, including those with congenital anomalies which include mandibular underdevelopment.⁴⁶ Yet only a relatively small proportion of the cases we studied reported prior polysomnography. These results suggest that physicians may not consistently address sleep-related problems in this population even when children snore frequently, and that SDB may be underdiagnosed among children with HFM.

Conclusion

In this follow-up study, snoring and symptoms of SDB were more commonly reported in children with versus without HFM. Among children with HFM, nearly 30% reported snoring and 23% reported symptoms suggestive of SDB. Because interventions for SDB are available, its identification via routine screening in children with HFM is important not only to treat symptoms of SDB, but also to alleviate its health- and behavior-related sequelae.

Abbreviations

adjOR

Adjusted odds ratios

AT

Adenoidectomy or tonsillectomy

HFM

Hemifacial microsomia

CI

95% Confidence interval

OR

Odds ratio

PSQ

Pediatric Sleep Questionnaire

SDB

Sleep-disordered breathing

PSQ+

Positive screen for sleep-disordered breathing based on the Pediatric Sleep Questionnaire