A diagnostic approach to suspected obstructive sleep apnea in children

ORIGINAL ARTICLES

A diagnostic approach to suspected obstructive sleep apnea in children Most children with obstructive sleep apnea will benefit from tonsillectomy and adenoidectomy. Although polygraphic monitoring remains the definitive diagnostic technique, we wondered i f all children suspected o f having OSA require such evaluation. We therefore administered a standardized questionnaire to the parents o f 23 children with polygraphically proved OSA related to adenotonsillar hypertrophy, 46 age- and sex-matched normal children, and 23 children subsequently referred because of possible OSA. Significantly increased frequencies o f the following symptoms were found in the OSA group compared with the control group: difficulty breathing during sleep, 96% vs 2%; apnea observed by the parents, 78% vs 5%; snoring, 96% vs 9%; restless sleep, 78% vs 23%; chronic rhinorrhea, 61% vs 11%; and mouth breathing when awake, 87% vs 18%. Using discriminant analysis, an OSA score was derived that correctly classified all control subjects and 22 o f 23 patients with OSA. Considering the data from all groups, we found that (I) OSA scores >3.5 were highl~ predictive o f OSA requiring adenotonsillectomy; (2) no child with an OSA score <-1 had OSA; and (3) in children with OSA scores between - I and 3.5, polygraphic monitoring was required to determine the severity o f sleep-related airway obstruction and the need for surgical treatment. Use o f the OSA score should decrease the need for polygraphic monitoring and facilitate selection o f children for tonsillectomy and adenoidectomy. (J PEDIATR 105:10, 1984)

Robert Brouilette, M.D., Donna Hanson, B.S.N., Richard David, M.D., Linda Klemka, B.A., Anna Szatkowski, C.R.T.T., Sandra Fernbach, M.D., and Carl Hunt, M.D. Chicago, Ill.

THE SYMPTOMS OF SLEEP-RELATED upper airway obstruction or obstructive sleep apnea have been described.t-~' In children, the most common cause of OSA is adenotonsillar hypertrophy? ,2,4-8 Unfortunately, the clinical usefulness of historical data has been limited because little information is available on the frequencies of symptoms in normal children? We therefore planned this study to define the frequencies of symptoms in the OSA population and in normal children. Use of such comparative data allows us to classify children into three groups: normal children with no significant symptoms who need no further evaluation; children with characteristic symptoms who require tonsillectomy and adenoidectomy; and children From the Department o f Pediatrics and Radiology, Children's Memorial Hospital and Northwestern University. Supported in part by the Children's Research Guild. Reprint requests: Robert T. Brouillette, M.D., Division o f Neonatology, Children's Memorial Hospital, 2300 Children's Plaza, Chicago, IL 60614.

10

The Journal o f P E D l A T R l C S

with nondiagnostic symptoms who warrant definitive evaluation by polygraphic monitoring.

METHODS A standardized questionnaire based on our clinical experience and published signs and symptoms of OSA was developed in early 1981. This questionnaire was completed for each patient or control by one investigator (D.H.), who OSA

Obstructive sleep apnea

interviewed one or both parents. The form was always completed before laboratory evaluation of patients referred for possible OSA. The questionnaire consisted of 57 questions that we initially suspected might help to distinguish patients with OSA from normal children, These questions included the following general areas: (1) signs of sleep-related upper airway obstruction; (2) sleep habits; (3) parenterally perceived abnormalities of behav-

Volume 105 Number 1

ior, learning, and development; (4) past history of ear, nose, or throat disease; and (5) family history of breathing difficulties, OSA, or tonsillectomy/adenoidectomy. For most questions pertaining to sleep-related breathing problems, the parents were asked to estimate the frequency with which their child demonstrated the problem; for example, parents were asked whether their child never, occasionally, frequently, or always snored during sleep. Most other questions required yes or no answers. The standardized history form usually took about 15 minutes to complete. Definitions. Previous authors have usually applied Guilleminault's definition of O S A in adults to children: " A sleep apnea syndrome is diagnosed if, during 7 hours of nocturnal sleep, at least 30 apneic episodes are observed in both REM and non-REM sleep, some of which must appear repetitively in non-REM sleep. ''~~ As detailed in our previous reports, 6`tt~3 we diagnosed OSA in children (1) when episodes of partial or complete airway obstruction during sleep resulted in abnormal tcPo2 (<50 mm Hg) or PACO2 (>45 mm Hg), and (2) when sleep-related asphyxia and sleep deprivation resulted in clinically significant effects such as failure to thrive, cor pulmonale, or neurobehavioral disturbances. This latter approach to the diagnosis of OSA thus combines the information gained from noninvasive estimates of blood gas abnormalities during sleep with the data from the clinical evaluation and other laboratory tests. Laboratory evaluations. In all children in the OSA and possible OSA groups, electrocardiograms were taken to rule out cor pulmonale, ~2and endolateral neck radiographs were taken to evaluate adenoidal and tonsillar size and to rule out other causes of airway obstruction, u In these two groups, polygraphic monitoring was performed during a daytime nap after sleep deprivation. Heart rate, ECG, tcPo2, PACO2, oral and nasal airflow, and respiratory motions of the thorax and abdomen were recorded on separate polygraph channels. 6' t3 Patient selection. The OSA group consisted of all children evaluated in our Sleep Laboratory from March 1981 to May 1982 who met the criteria for study. Because we were trying to develop a screening test for OSA related to adenotonsillar hypertrophy, we excluded children with prior tonsillectomy or adenoidectomy, known neurologic disease, or craniofacial abnormalities. Twenty-three children with polygraphically proved OSA met these criteria. All 23 were between 1 and 10 years of age, and all had moderate to marked adenoidal or tonsillar enlargement on endolateral neck radiographs as interpreted by a pediatric radiologist (S.F.). n The control group was obtained from the general pediatric clinics at Children's Memorial Hospital and from a general pediatric office practice. Because of previous

Suspected obstructive sleep apnea

11

experience suggesting that OSA in children is related to age and sex, each patient with OSA was matched with two controls of the same sex and age (within 1 year). The possible OSA group consisted of 23 patients subsequently (May 1982 to January 1983) referred to the Children's Memorial Hospital Sleep Laboratory for possible OSA who met the criteria for study. The purpose of including the possible OSA group was to determine whether a discriminant function based on this questionniare could predict whether a child referred for possible OSA would or would not have OSA. During this same interval, 21 other children referred for possible OSA were excluded from analysis because of airway abnormalities other than adenotonsillar hypertrophy (nine patients), age <1 year (five), neurologic disease (five), or prior adenoidectomy or tonsillectomy (five). Data analysis. We analyzed our data in three steps. First, the responses to each questionnaire item for the OSA and control groups were compared for group differences. Chi-square analyses were used for nominal and ordinal variables; two-tailed t tests were used for continuous variables. Second, a discriminant function, the OSA score, was derived that allowed classification of individuals into the OSA or control groups by considering their responses to combinations of several questions.~4. ~5Although increasing the number of predictor variables in a discriminant function will increase the classifying power for those cases from which the function was actually derived, this procedure can lead to more misclassification when cases are taken from a different sample. Thus our strategy was to use as few variables as possible to achieve good discrimination, thus improving the expected predictive power in a new population. Third, the OSA score was evaluated to determine how well it predicted the diagnosis of OSA (as defined physiologically) in a new group of patients, the possible OSA group. Stepwise discriminant function analysis was performed on the known OSA and control groups using the 11 variables that performed best (i.e., had chi-square or t test results with P < 0.005) in the earlier, univariate analysis. The f-to-enter value used was 2.0. Discriminant functions with three to six variables were then further evaluated by randomly splitting the cases and controls into two subsamples and observing the stability and external validity of the functions derived on these subsets. RESULTS

Single-variable comparisons of OSA and control groups. The age, sex, and race distributions were similar for the OSA and control groups. There was a preponderance of boys (16 of 23) and preschool aged children in the OSA

12

B r o u i l l e t t e et al.

The Journal o f Pediatrics J u l y 1984

Table. Frequency of signs and symptoms in OSA and control groups OSA group Sign~symptom

Difficulty breathing when asleep* Stops breathing during sleep Snoring* Restless sleep* Chronic rhinorrhea Mouth breathes when awake Frequent upper respiratory tract infections Frequent nausea/vomiting Difficulty swallowing Sweating when asleep Hearing problem Excessive daytime somnolence Poor appetite Recurrent middle ear disease Pathologic shyness/social withdrawal Delayed development Nightmares* Speech problemst History of pneumonia Increased difficulty breathing during upper respiratory tract infections Frequent morning headachest Nocturnal enuresist Clumsy/lack of coordination Aggressive Hyperactive* Major discipline problem Bizarre behavior

Control group

n

%

n

%

22/23 18/23 22/23 18/23 14/23 20/23 19/23 7/23 6/23 11/22 3/23 7/21 7/23 10/23 5/23 4/23 3/23 4/13 7/23 20/23

96 78 96 78 61 87 83 30 26 50 13 33 30 43 22 17 13 31 30 87

1/45 2/41 4/45 10/44 5/46 8/44 13/46 1/46 1/46 7/45 0/44 4/44 4/46 8/46 2/44 2/46 0/44 3/28 8/46 29/38

2 5 9 23 11 18 28 2 2 16 0 9 9 17 5 4 0 11 17 76

0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.002 0.007 0.014 0.014 0.019 0.019 0.027 0.066 0.067 0.11 0.213 0.312

2/13 1/ 13 5/23 4/23 4/13 2/23 3/22

15 8 22 17 31 9 14

2/28 1/26 8/46 16/46 7/27 6/46 2/45

7 4 17 35 26 13 4

0.41 0.61 0.607 0.129 0.75 0.943 0.175

*For these questions,"'constant"and "'frequent"responseswere regarded as "yes," +'occasional"or "never" were regarded as "no.'" +For these symptoms,onlychildren>_3 years of age were considered. group. The mean ages of the OSA and control groups were 3.8 +_ 2.4 (SD) and 4.0 _+ 2.3 (SD) years, respectively. Initial inspection of the parental responses to questions related to sleep and breathing showed that OSA group parents usually responded "constantly" or "frequently" to these questions, whereas control group parents usually responded "occasionally" or "never" (Table). These four categories were therefore collapsed "into "yes" and "no" responses, respectively. The OSA group had a much higher incidence of snoring, difficulty breathing during sleep, sweating during sleep, and restless sleep. Seventy-eight percent of the OSA group parents, but only 4% of control group parents, stated that they had seen their child stop breathing during sleep. The OSA group slept less at night but more during the day than the control group children. The average nocturnal sleep durations were 8.7 _ 2.2 (SD) hours in the OSA group and 10.2 _+ 1.4 (SD) hours in the control group (P < 0.001). The OSA group had an increased incidence of several signs suggesting adenoidal hypertrophy: chronic rhinorr-

hea, recurrent middle ear disease, hearing problems, mouth breathing when awake, and frequent upper respiratory tract infections. The OSA group also had an increased incidence of poor appetite and difficulty swallowing (Table). The responses to questions about behavior, learning, and development were not markedly different for the OSA and control groups. The increased incidence of "pathology shyness/social withdrawal," "delayed development," and "nightmares' 'in OSA children was of marginal significance considering the large number of variables evaluated. There were no significant differences in the family histories of breathing difficulties, OSA, adenoidectomy, tonsillectomy, sudden infant death syndrome, or infantile apnea. Possible OSA group. The 23 patients subsequently referred for OSA evaluation were slightly older (5.3 _+ 3.6 (SD) years, P < 0.05) than the other two groups. Seventeen (74%) of the 23 children were boys. The final diagnoses in these patients, made after history, physical examination during sleep, endolateral neck radio-

Volume 105 Number 1

Suspected obstructive sleep apnea

13

6 20"

03 15-

3 lO-

z= 5 0

'

-4

i

-3

.

i .

-2

.i

-I

[] II i.

OSA

i

.

0

i.

i

i

i

I

2

3

4

7=

-4

i -3

SCORE

Fig. 1. Distribution of OSA scores in known OSA (11) and normal control groups (n). OSA scores <0 predict normality; values >0 predict OSA. Note that only one patient with OSA was misclassified and all normal controls were classified correctly.

graphs, ECG, and polygraphic monitoring, were normal in eight children, OSA in 10, and borderline OSA in five. Each of the patients diagnosed as having borderline OSA had chief complaints of loud snoring and sleep-related breathing difficulty. Each had moderate to marked adenotonsillar enlargement on endolateral neck radiographs, and each had normal ECG findings. On polygraphic monitoring, the only complete airway obstruction was one 8-second episode in one patient. However, each of these five patients showed some evidence of partial sleep-related airway obstruction: highest PAco2 during sleep ranged from 44 to 48 mm Hg; lowest tcPo2 during sleep ranged from 48 to 70 mm Hg; and the percent of sleep time with paradoxical inward chest movement on inspiration ranged from 6% to 64%. Diseriminant analysis. Stepwise discrimination analysis in the OSA and control groups yielded a six-variable equation that explained 94% of the variance in final diagnosis. Because all but 1% of this variance came from the first three variables, we used the following threevariable function: OSA score = 1.42D + 1.41A + 0.71S - 3.83 where D is difficulty breathing during sleep, A is apnea observed during sleep, and S is snoring. Values assigned to D and S were 0, never; 1, occasionally; 2, frequently; and 3, always. Values assigned to A were 0, no; and 1, yes. The OSA score correctly identified all controls and correctly assigned all but one of the OSA group (Fig. 1). The OSA score was tested in three ways. First, when the OSA score was derived from one half the cases and used to predict diagnosis for the other half, 91% and 97% of cases, respectively, were correctly classified. Second, the OSA score was calculated for each of the children with possible OSA. (Fig. 2). The OSA score correctly predicted three of

-2

-I

OSA

0

I

2

3

4

SCORE

Fig. 2. Distribution of OSA scores by final diagnosis for possible OSA cases. OSA scores <0 predict normality; scores >0 predict OSA. Note that one of eight patients finally diagnosed as having OSA (11) and four of seven diagnosed as normal (I-1) were misclassified using OSA score. Patients diagnosed as having borderline OSA ([]) are also shown.

seven normal children and eight of nine with OSA. Three of five borderline cases were predicted to have OSA. Third, linear regressions of the OSA score on the lowest tcpo2, the highest PAr and the percentage of sleep time with inward chest movement on inspiration for the known OSA and possible OSA groups gave correlation coefficients of -0.37, 0.41, and 0.38 (all P < 0.01), respectively. Thus, higher OSA scores predicted lower tcPo2, higher PAr and increased percentage of sleep time with inward chest movement on inspiration. DISCUSSION The frequencies of symptoms in our normal and OSA groups are similar to those in previous reports studying either of these populations. Richardson et al# also found a high incidence of snoring (100%) and apnea observed by the parents of children with OSA (60%). Frank et al.* found snoring in 91% and apnea observed by parents in 81% of children with polygraphically proved OSA. Our data may actually underestimate symptom frequency in children with OSA, because the most obvious cases were probably refeferred directly to surgeons for tonsillectomy and adenoidectomy. In a sample of 355 normal children, Weissbluth et al. 9 found that 27% snored and 6% had difficulty breathing during sleep. These percentages may be slightly higher than comparable percentages in our control group, because we included as positive responses only those children who frequently or constantly snored and had difficulty breathing during sleep. Thus, in a general office practice, a history of frequent snoring and difficulty breathing during sleep or of obstructive apnea observed by the parents strongly suggests that the child has OSA.

14

Brouillette et al.

At least two factors may have contributed to the failure of our OSA score to predict which patients subsequently referred for evaluation would have OSA. First, discriminant functions generally perform better on the sample from which they are derived than they do when applied to a new sample. ~4 This factor is probably not important, because the OSA score worked quite well when derived from half the OSA and control cases and used to predict diagnosis in the other half. The second and more important factor relates to the variable severity of sleep-related upper airway obstruction in children. The most affected patients show frequent complete airway obstruction during sleep and severe gas exchange abnormalities (PACO2> 60 mm Hg, tcPo2 < 20 mm Hg). These patients would meet the usual definition for OSA in adults. 1~ Most children with OSA are not quite so severely affected, having relatively few episodes of complete airway obstruction but frequent partial airway obstruction and moderately abnormal gas exchange (PAco2 25 to 60 mm Hg, tcPo2 20 to 50 mm Hg). The next most affected group we called borderline OSA: these patients had clinical symptoms, borderline gas exchange values (PAr 45 mm Hg, tcPo~ 50 mm Hg), and intermittent inward chest movement on inspiration. Finally, the least affected patients had clinical symptoms such as snoring and difficulty breathing during sleep but normal findings on polygraphic monitoring. These latter two groups may be analogous to adult snorers who have been shown to be physiologically intermediate between normal subjects and adults with OSA. ~6.~7 Thus, in our possible OSA group, patients with a final diagnosis of OSA, borderline OSA, or "normal" may all have had some degree of sleep-related airway obstruction. Insufficient data exist at present to determine whether adenotonsillectomy would be beneficial in the borderline OSA and "symptomatic normal" children. In which children suspected of having OSA resulting from adenotonsillar hypertrophy is polygraphic monitoring warranted before adenoidectomy or tonsillectomy? The data in the Figures may be interpreted as follows: (1) an OSA score >3.5 is diagnostic of OSA; (2) an OSA score < - 1 indicates absence of OSA; and (3) OSA scores between - 1 and 3.5 may or may not indicate OSA. Thus we suggest that polygraphic monitoring is not necessary in those children who meet our selection criteria and who have OSA scores >3.5 or < - 1 . Polygraphic monitoring should be used in children with suspected OSA if the OSA score is between - l and 3.5 or if confounding factors, such as neurologic disease or craniofacial abnormalities, are present. We thank the Ambulatory Pediatrics Division at Children's Memorial Hospital and Drs. Howard Rice and Mitchell Blivaiss

The Journal of Pediatrics July 1984

for allowing us to interview their patients' parents; Drs. Gabriel Tucker, Carol Gerson, Lauren Holinger, and Marc Weissbluth for review of the manuscript; and Ms. Susan Seidler and Ms. Judy Carbone for preparation of the manuscript,

REFERENCES

1. Guilleminault C, Eldridge FL, Simmons FB, Dement WC: Sleep apnea in eight children. Pediatrics 58:23, 1976. 2. Konno A, Hoshino T, Togawa K: Influence of upper airway obstruction by enlarged tonsils and adenoids upon recurrent infection of the lower airway in childhood. Laryngoscope 90:1709, 1980. 3. Rowe LD, Hansen TN, Nielson D, Tooley WH: Continuous measurements of skin surface oxygen and carbon dioxide tensions in obstructive sleep apnea. Laryngoscope 90:1797, 1980. 4. Eliasehar I, Lavie P, Halperin E, Gordon C, Alroy G: Sleep apneie episodes as indications for adenotonsillectomy. Arch Oto[aryngol 106:492, 1980. 5. Guilleminault C, Korobkin R, Winkle R: A review of 50 children with obstructive sleep apnea syndrome. Lung 159:275, 1981. 6. Brouillette RT, Fernbach SK, Hunt CE: Obstructive sleep apnea in infants and children. J PEDIA'rR100:31, 1982. 7. Richardson MA, Seed AB, Cotton RT, Bento C, Kramer M: Evaluation of tonsils and adenoids in sleep apnea syndrome. Laryngoscope 90:1106, 1980. 8. Frank Y, Kravath RE, Pollak CP, Weitzman ED: Obstructive sleep apnea and its therapy: Clinical and polysomnographic manifestations. Pediatrics 71:737, 1983. 9. Weissbluth M, Davis AT, Poncher J, Reiff J: Signs of airway obstruction during sleep and behavioral, developmental and academic problems. J Dev Behav Pediatr 4:119, 1983. 10. Guilleminault C, Tilkian A, Dement WC: The sleep apnea syndromes. Ann Rev Med 27:465, 1976. 11. Fernbach SK, Brouillette RT, Riggs TW, Hunt CE: Radiologic evaluation of adenoids and tonsils in children with obstructive sleep apnea: Plain films and fluoroscopy. Pediatr Radiol 13:258, 1983. 12. Hunt CE, Brouillette RT: Abnormalities of breathing control and airway maintenance in infants and children as a cause of cor pulmonale. Pediatr Cardiol 3:249, 1982. 13. Jeffries B, Brouillette RT, Hunt CE: Electromyography of accessory muscles of respiration in children referred for obstructive sleep apnea (OSA). Am Rev Respir Dis, April 1984. 14. Kleinbaum DG, Jupper LL: Applied regression analysis and other multivariable methods. North Scituate, Mass., 1978, Duxbury Press, pp 414-447. 15. Klecka WR: Discriminant analysis. In Nie NH, Hull CH, Jenkins JG, Steinbrenner K, Bent DH, editors: Statistical package for the social sciences, ed 2. New York, 1975, McGraw-Hill, pp 434-467. 16. Lugaresi E, Coccagna G, Farneti P, Mantovani M, Cirignotta F: Snoring. Electroenceph Clin Neurophysiol 39:59, 1975. I7. Lugaresi E, Cirignotta F, Coccagna G, Piana C: Some epidemiological data on snoring and cardiocirculatory disturbances. Sleep 3:221, 1980.