Archives ofDisease in Childhood 1993; 68: 360-366

Snoring, sleep disturbance, and behaviour in45 year olds

N J Ali, D J Pitson, J R Stradling

AbstractParents of 996 children aged 4-5 years identi-fied consecutively from the Oxford healthvisitor register were asked to complete aquestionnaire about breathing disordersduring sleep. A total of 782 (78.5%) wasreturned. Ninety five (12-1%) children werereported to snore on most nights. Habitualsnoring was significantly associated with day-time sleepiness, restless sleep, and hyper-activity.The questionnaire responses were used to

select two subgroups, one at high risk of a sleepand breathing disorder and a control group.These children (132 in total) were monitored athome with overnight video recording andoximetry, and had formal behavioural assess-ment using the Conners scale.Seven (7/66) children from the high risk

group and none from the control group hadobvious sleep disturbance consequent on snor-ing and upper airway obstruction. Thus ourestimate of the prevalence of sleep and breath-ing disorders in this age group is 7/996 or 07%.The high risk group had significantly higher

nocturnal movement, oxygen saturation diprates, and overnight pulse rates than the con-trols. Maternal but not paternal smoking wasassociated with the high risk group. Parentsand teachers thought those in the high riskgroup were more hyperactive and inattentivethan the controls, but only their parentsthought them more aggressive.

Significant sleep and breathing disordersoccur in about 07% of 4-5 year olds. Childrenwhose parents report snoring and sleep dis-turbance have objective evidence of sleepdisruption and show more behaviour problemsthan controls.(Arch Dis Child 1993; 68: 360-366)

Osler Chest Unit,Churchill Hospital,Headington, OxfordN J AliD J PitsonJ R StradlingCorrespondence to:N J Ali, Department ofRespiratory Medicine,Addenbrooke's Hospital, HillsRoad, Cambridge CB2 2QQ.Accepted 21 October 1992

Sleep and breathing disorders are increasinglyrecognised in children and are usually caused bya combination of adenotonsillar hypertrophy anddecreased oropharyngeal dimensions.' There hasbeen a dramatic decrease in tonsillectomy ratesin children up to 9 years ofage from about 40% inthe 1950s2 to less than 1% by the same age in theyears 1975-80.3 This decreasing trend, and theoften non-specific and poorly recognised symp-toms associated with sleep and breathing dis-orders in children,4 has led to the suggestion thatthe prevalence and consequences of these sleeprelated problems is increasing.5 The indicationsfor adenotonsillectomy have also changed; one

study found that obstructive sleep apnoea, whichwas not reported in 1978, accounted for 19% ofthe procedures in 1986.6

This study has therefore addressed questionsabout the prevalence and psychological impact ofundiagnosed sleep and breathing disorders inchildren: first, through a questionnaire surveyabout snoring and sleep problems in 996randomly selected children aged 4 5-5 years;and second, by evaluating with home sleepstudies and formalised behavioural assess-ments a high risk subgroup and comparingthese children with a low risk control subgroup,with the two groups selected from the initialsurvey.

SubjectsQUESTIONNAIRE SURVEYNine hundred and ninety six boys and girlsborn consecutively between 8 January and 22September 1985 and living in Oxford wereidentified from the centralised health visitorregister. The ages at the time of study rangedfrom 4- 5 to 5 * 5 years.

SLEEP STUDY AND BEHAVIOURAL ASSESSMENTGROUPOn the basis of the answers to the initialquestionnaire, two subgroups were selected forfurther study. Questions 1, 5, 6, 9, 10, and 11(essentially about snoring and other symptoms ofsleep and breathing disorders; see appendix)were given a weighting based on their ability topredict the presence of sleep and breathingdisorders in a previous study from this depart-ment.7 The overall score was then used to rankthe children in order of the likelihood of themhaving a sleep and breathing disorder. The 66highest ranking children (henceforth the highrisk group) and an equal number of sex matchedchildren from the lowest rankings (henceforththe control group) were studied further. Theirparents were contacted by letter and telephoneand asked if they would allow their children totake part in the study. Of 73 approached in thehigh risk group 66 (90%) agreed to the study,whereas we contacted 88 in the control group toobtain an equal number to study (75%). Thus132 children from the original target populationof 996 (13-3%) had sleep studies and psycho-logical assessment.

MethodsQUESTIONNAIRE SURVEYThe postal questionnaire inquired about snoringand other symptoms of sleep and breathingdisorders in children (see Appendix) and was areduced version of that used in an earlier studyfrom this department.7

360

on January 22, 2021 by guest. Protected by copyright.

http://adc.bmj.com

/A

rch Dis C

hild: first published as 10.1136/adc.68.3.360 on 1 March 1993. D

ownloaded from

Snoring, sleep disturbance, and behaviour in 4-5year olds

SLEEP STUDY AND BEHAVIOURAL ASSESSMENTGROUPSGeneral health, enuresis, parental smoking, andoccupation questionnaireWe inquired about the child's general health(particularly hearing problems), drugs taken,and whether they had regular enuresis. Thesocioeconomic group was determined frompaternal occupation according to the Office ofPopulation and Censuses and Surveys (OPCS)definitions,8 and parental smoking habits werealso determined as these factors might be poten-tially confounding variables in relating snoringand behavioural assessment.9

Overnight video recordingsA low light video camera, infrared light source,microphone, and video recorder were set up inthe child's bedroom to run between 10 pm and6 am. The recordings were processed automatic-ally for body movements as a measure of sleepdisturbance70 which was expressed as the per-centage of time spent moving. All video record-ings were reviewed blind before the estimation ofthe percentage of time spent moving to removeperiods of longer wakefulness, thus ensuringthat this index was a measure of the transientarousals characteristic of sleep and breathingdisorders. Any snoring and obstructed breathingproducing arousals was documented to classifythe children as normal, snorers, or snorers/sleepapnoea with sleep disturbance.

OximetryThe Ohmeda 3700 pulse oximeter with an eighthour internal memory was used to make over-night recordings.1" By asking the parents torecord accurately the time the oximeter wasswitched off in the morning we were able tosynchronise the video and oximeter recordings towithin a minute or so. The oximetry data wereunloaded into a BBC microcomputer and theaverage oxygen saturation and pulse rate calcu-lated. The number of dips in arterial oxygensaturation (Sao2) greater than 4% was also calcu-lated and expressed as >4% dips per hour (>4%Sao2 dip rate) as described previously.'2

Conners behaviour rating scaleThe Conners scale was shortened by includingonly the three subscales (out of a normal five)that related to the behaviour patterns reported inassociation with sleep apnoea in children (that is,aggression, inattention, and hyperactivity). 13We used the subscales and scoring system asreported by Conners. 13

ProtocolQUESTIONNAIRE SURVEYThe questionnaire, together with a letter thatindicated our interest in children's sleep ingeneral (not specifically in sleep and breathingdisorders), was sent to the parents of the 996children identified from the health visitorrecords. If no reply was received within three tofour weeks one reminder was sent.

SLEEP STUDY AND BEHAVIOURAL ASSESSMENTGROUPSThe two subgroups (132 children) identifiedfrom the initial questionnaire answers werevisited at home and their parents completed thegeneral health and parental smoking and occupa-tion questionnaire. The sleep study equipmentwas set up in the child's bedroom and collectedthe next day. If the child appeared to have acurrent upper respiratory tract infection, thenthe sleep study was deferred until they hadrecovered. At a later date (within three months)the parents were asked to complete the Connersbehaviour scale themselves and to pass anothercopy onto the child's teacher for completion.

AnalysisAll the data were entered into the OxfordUniversity VAX computer and analysed usingthe SAS suite of statistical programs'4 or theGeneral Linear Interactive Modelling (GLIM)statistical package'5 as appropriate.

QUESTIONNAIRE SURVEYThe questionnaire responses were examined forassociations between reported snoring anddaytime sleepiness, hyperactivity, and restlesssleep. The original four possible responses tothe snoring question were retained and givenan arbitrary score of 1= never, 2=rarely, 3=sometimes even without a cold, and 4=mostnights, whereas those for daytime sleepiness,hyperactivity, and restless sleep were collapsedinto two categories: never/rarely and sometimes/often. Linear trends across the snoring categor-ies in daytime sleepiness, hyperactivity, andrestless sleep were tested for statistical signific-ance using X2 trend analysis.'6 The relative risksand 95% confidence intervals were calculated forhabitual snorers (that is, most nights) comparedwith non-snorers (that is, never) using the confi-dence interval analysis computer program. 11

SLEEP STUDY AND BEHAVIOURAL ASSESSMENTGROUPSDetermination ofthe prevalence ofsleep andbreathing disordersThere are no accepted criteria for the diagnosisof sleep and breathing disorders in childrenbecause there is a continuum from normality togross disease with uncertain correlations andconsequences. We chose to use the 95th centileestablished for a group of normal children (pre-viously reported from this unit7 which utilisedidentical techniques) to define the upper limit ofnormality for nocturnal movement (percentageof time spent moving), >4% Sao2 dip rate, andmean overnight pulse rate. The reviews of thevideo and oximetry records of these children(who were above the 'normal' range for one ormore of the measurements) were then used todetermine whether a sleep and breathing dis-order or another factor was the likely explanationfor their abnormal result.

Comparison ofhigh risk and control groupsOvernight sleep monitoring. The two groups

361

on January 22, 2021 by guest. Protected by copyright.

http://adc.bmj.com

/A

rch Dis C

hild: first published as 10.1136/adc.68.3.360 on 1 March 1993. D

ownloaded from

Ali, Pitson, Stradling

Table 1 Effect ofsnoring on prevalence ofdaytime sleepiness, hyperactivity, and restless sleep (n= 782). Relative risk comparesthe habitual snorers to the never snorers

Snoring catego?y95%

Never Rarely Sometimes Most nights x2 trend Relative Confidence(n=247) (n=239) (n=201) (n=95) (p value) risk interval

Daytime sleepiness 19 17 26 31 3-88 1-18 1-02 to 1-36(%) (p=0 048)

Hyperactivity (%) 19 23 31 52 31-9 1-74 1-38 to 2-18(p<o0oool)

Restlesssleep((%) 30 36 50 64 39-3 2 06 1-53to2-76(p<O-0001)

were compared with respect to the percentage oftime spent moving, >4% Sao2 dip rate, and pulserate using an unpaired t test.

Social class and parental smoking. For thepurposes of analysis the six socioeconomiccategories defined by the OPCS were collapsedinto two: manual - that is, groups Illm, IV and V- and non-manual - that is, groups I, II, andIIInm. Single logistic regression was used tocalculate the effect of social class and parentalsmoking on the odds ratio for being in the highrisk group. The odds ratios were then recalcu-lated in a logistic regression model after allowingfor potentially confounding variables - forexample, for social class we entered maternal andpaternal smoking into the analysis first. 15

Conners child behaviour scale. To producetwo categories to be used as the dependantvariable in logistic regression we used the 95thcentile for the control group to define the upperlimit of 'normality' in our study. Thus anychildren whose scores equalled or exceeded thisvalue were defined for the purposes of analysis as'abnormal'. As in the social class and parentalsmoking analysis single logistic regression wasused to determine the odds ratio effect ofbeing inthe high risk versus the control group on thescore for each of the three subscales. The oddsratios were then recalculated in a logistic regres-sion model after controlling for gender and socialclass as potentially confounding variables.

ResultsQUESTIONNAIRE SURVEYSix hundred and seven (607/996; 60 9%) com-pleted questionnaires were returned initially anda further 175 after the reminder letter (totalreturned 78 5%). Of these, 417/782 (53 3%)were boys. Thirty per cent of the parentsreported their children as having problemsrelated to sleeping sometimes or often, 22-3%

Table 2 Summary ofresults ofsleep monitoring in the two study groups, showing measurementin which they exceeded the normal range and categorised by whether they had a sleep andbreathing disorder (SBD), snored, or had apparently norrnal sleep when the video recording wasreviewed

High risk group Control groutpSBD Snorer 'Normal' Sniorer 'Nornall

>4% Sao2 dip rate 2 4 5Nocturnal movement 2 - 3* 4t*Pulse rate - I>4"YO Sao. dip rate and movement 3 - - - 2tNormal - 11 35 1 59Total 7 15 44 1 6

*Child with eczema.tChild with upper respiratory infection on studsv night.

sometimes or often fell asleep while reading orwatching television, and 27-3% considered theirchildren to be hyperactive. A total of 12y1%snored on most nights and 25 7% snored often,even without a cold. Restless sleep was reportedby 31-1% and mouth breathing by 17 5% ofparents, whereas 9 8% remembered being told atone time that their child had enlarged tonsils.Nine children had had a tonsillectomy, four ofthese had by chance been investigated in thisdepartment between one and one and a half yearsearlier and found to have a clear cut sleepand breathing disorder before their tonsil-lectomy.

Questionnaire interrelationsTrend analysis (x2) showed the increasing preval-ence of daytime sleepiness, hyperactivity, andrestless sleep across the four snoring categories toall be significant (table 1). Daytime sleepiness,hyperactivity, and restless sleep were all signific-antly more common among the habitual snorers(that is, most nights) compared with the neversnorers (table 1).

SLEEP MONITORING AND BEHAVIOURALASSESSMENT GROUPSSleep monitoring resultsOvernight video recordings were refused ortechnically unsatisfactory in 12/132 children(seven from the high risk group), and oximetrvyrecordings were not available for the samereasons in another 15 children (seven from thehigh risk group). Table 2 summarises the resultsof home sleep monitoring.

In the high risk group 20/66 (30%) childrenhad values for nocturnal movement, >4% Sao,dip rate, or overnight pulse rate greater than the95th centile for normal children. In seven therewas clear cut snoring with recurrent upperairway obstruction and consequent sleep dis-turbance. Another 15 children snored on thenight of the study and four ofthem had abnormaloximetry records, but review of the video record-ing did not show snoring related sleep disturb-ance and so they could not be categorised ashaving a sleep and breathing disorder. One childhad eczema with scratching, which explained thehigh nocturnal movement. Review of theremaining eight children with abnormaloximetry, pulse rate, or movement recordsshowed apparently normal sleep with no snoringor other explanation for their results.

In the control group 6/66 (9%) of these child-ren had abnormal overnight monitoring results.This could be explained on the basis ofeczema in

362

on January 22, 2021 by guest. Protected by copyright.

http://adc.bmj.com

/A

rch Dis C

hild: first published as 10.1136/adc.68.3.360 on 1 March 1993. D

ownloaded from

Snoring, sleep disturbance, and behaviour in 4-5year olds

Table 3 Results ofovernight sleep monitoring in the high riskand control group (excluding the seven children with anidentifiable sleep and breathing disorder)

Control High riskgroup group p Value(n=58) (n=52)

% Time spent moving (SD) 4 9 (2 3) 5 8 (1-9) 0 043>4% Sao2 dip rate (SD) 1-4 (0 8) 2-1 (1-4) 0 0006Mean overnight pulse rate (SD) 81-0 (7 2) 83-8 (7 2) 0-031

one and upper respiratory infection in anothertwo children (which had not been apparentduring the interview the previous evening). Theother three had apparently normal sleep. Onlyone child (1/66, 2%) in this group snored on thenight of monitoring and her results were withinthe normal range.

Prevalence ofsleep and breathing disordersWe identified seven children with a detectablesleep and breathing disorder during this studyfrom our original population of 996 children, aprevalence of 0-7%. If we include the fourchildren from this population who had beenidentified previously (see questionnaire resultssection) with a sleep and breathing disorder andwho are unlikely to have improved withoutadenotonsillectomy then our estimate of the pre-valence of these disorders is even higher at 1I%(7+ 4/996). This is likely to be an underestimateas we studied only 16-8% of the original popula-tion.

Comparison ofhigh risk and control groupsInitial analysis of the overnight sleep monitoringresults showed that the high risk group hadsignificantly higher nocturnal movement, >4%Sao2 dip rate, and pulse rate than the controls,but this included the seven children identifiedwith an obvious sleep and breathing disorder(who by definition had high nocturnal move-ment, >4% Sao2 dip rate, and pulse rate). Theseseven children were therefore subsequentlyexcluded from the comparison of the high riskand control groups. The high risk group never-theless had significantly higher nocturnal move-ment, >4% Sao2 dip rate, and overnight pulserate (table 3). Thus, even though these childrendid not have a significant sleep and breathingdisorder by our criteria, their separation accord-ing to the questionnaire results reflected smallbut measurable physiological differences insleep.

Table 4 Relationships between high risk (n=66) and control (n=66) groups and social classand parental smoking

Odds ratio and 95% confidence interval

Questionnaire ranking After allozwanceBefore allowance for for other

Control group High risk group other variables varnables

Social class:Manual (%) 36 66 3-26 1-65Non-manual (%) 64 34 (1-60 to 6 65) (0-57 to 4 5)

Mother:Non-smoker (%) 77 54 2-99 4-36Smoker (%) 23 46 (I140 to 6 40) (I145 to 13l*)

Father:Non-smoker(%) 76 64 1 78 141Smoker(%) 24 36 (0-83 to 3-28) (0-48 to 2-51)

Children in the high risk group were signific-antly more likely to have a mother who smoked.This was independent of the effect of social class(table 3). Conversely, the increased likelihood ofbeing in the high risk group associated withcoming from a manual socioeconomic class waslost after allowing for the effect of maternalsmoking (table 4). The increased risk associatedwith paternal smoking was not significant. Therewas no difference between the two groups withrespect to the prevalence of regular nocturnalenuresis (11/66 in the high risk and 10/66 in thecontrol group).

Eighty six completed parent questionnaireswere returned and analysed (34/66 from the highrisk group and 52/66 from the control group),and 85 completed teacher questionnaires werereturned (35/66 from the high risk and 50/66from the controls).

Table 5 shows the median and 95th centilescores for each of the subscales. The highrisk group had higher scores for all the sub-scales except the teacher reported inattentivebehaviour. These data were transformed as des-cribed under Methods for use in logistic model-ling.

Social class appeared initially to be a signifi-cant predictor of parental Conners scores onunivariate analysis. After allowing for the effectsof high risk/control category first, however, it nolonger remained a significant independent pre-dictor. This was due to a relation between socialclass and high risk/control category, which inturn was through the maternal smoking effectreferred to earlier (table 4). Although social classdid not significantly affect the Conners scoresindependently, it was nevertheless allowed forfirst in calculating the subsequent odds ratios.

Although the groups overall were matched forgender, this was not so for the subset for whichwe had Conners results. As boys tend to scorehigher on the Conners scale9 than girls it wasallowed for as a possible confounding variable.

Table 6 gives the full logistic modellingresults. Children in the high risk group weresignificantly more likely to be thought inatten-tive and hyperactive by their parents andteachers. Aggressive behaviour was significantlymore likely to be reported by parents of the highrisk children. The trend towards increasedaggressive behaviour reported in the teacherquestionnaire for this group was not significant.Entering social class and gender into the modelfirst did not significantly change the odds ratiosfor any of the subscales.

DiscussionThis is the first community based study to reportthe prevalence of sleep and breathing disordersin children. None of the children we identified inthis study had severe obstructive sleep apnoea,but a less severe form of obstructive snoring(with some apnoeas) which nevertheless leads torecurrent sleep disruption (evidenced by highernocturnal movement), greater >4% Sao2 diprate, and higher pulse rate was seen. We havealso found significant differences in the resultsof sleep monitoring and daytime behaviourbetween children defined as being at high risk of

363

on January 22, 2021 by guest. Protected by copyright.

http://adc.bmj.com

/A

rch Dis C

hild: first published as 10.1136/adc.68.3.360 on 1 March 1993. D

ownloaded from

Ali, Pitson, Stradling

Table 5 Conners subscale scores for high risk and controlgroups as defined by questionnaire ranking; the median and 5thto 95th centiles are shown. The maximum score on anysubscale is 3

Conners subscale Control group High risk group

Aggressive behaviour (teacher) 0-06 (0-0 87) 0-12 (0-0 87)Aggressive behaviour (parent) 0-33 (0-1-16) 0-81(0-2-06)Inattentive behaviour (teacher) 0 50 (0-1-16) 0 50 (0-2 06)Inattentive behaviour (parent) 0-33 (0-1 16) 0 75 (0-2 16)Hyperactive behaviour (teacher) 0-06 (0-087) 0-13 (0-1 81)Hyperactive behaviour (parent) 0-14 (0-0 85) 0-42 (0-2 42)

a sleep and breathing disorder (according toparental history) compared with controls.

QUESTIONNAIRE RESULTSOur questionnaire was primarily designed toidentify children with sleep and breathing dis-orders for the second phase of the study. It wasderived from a questionnaire used in a previousstudy of children before a tonsillectomy.7 Thequestions which had proved most predictive ofthe presence of sleep and breathing disorderswere retained as well as a few others aboutdaytime behaviour. The usefulness of thequestionnaire results in defining a group at highrisk for sleep and breathing disorders is clearlyshown by the results of overnight monitoring,which tended to confirm parental history. Aresponse rate of 79% was achieved after onereminder; we did not pursue this further becauseour main concern was to find children with sleepand breathing disorders. It is likely that parentswho perceived their children as having a sleepproblem would be motivated to complete andreturn the questionnaire. This is supportedby the finding that parents were more likely toagree to the sleep study if their child was at highrisk of a sleep and breathing disorder (90%accepted compared with 75% of control subjectsapproached; see under Subjects). There is there-fore no reason to believe that there are significantnumbers of children with sleep and breathingdisorders among the non-responders. A total of12 -1% of children in our study were reported tosnore on most nights, a higher figure than thatreported by Corbo et al. 8 The children in ourstudy were considerably younger than theirsample and this may explain some of the differ-

Table 6 Distribution of 95th centile scores on Conners subscales according to whether the childwas in the high risk or control group. The odds ratio are calculated before and after allowingforgender and social class as possible confounding variables

Odds ratio and 95% confidence interval

Questionnaire ranking (%) After allowanceConners subscale Before allowance for for otherand centile Control group High risk group other variables variables

Aggressive behaviour (teacher)<95th 92 86 19 1 5>95th 8 14 (0 4to7-9) (0 3 to6-7)

Aggressive behaviour (parent)<95th 94 61 99 9.3>95th 6 39 (2-4 to 39-4) (2-3 to 39-1)

Inattentive behaviour (teacher)<95th 92 71 4-6 5 6>95th 8 29 (1-3 to 21-4) (1S to 21-4)

Inattentive behaviour (parent)<95th 94 73 5 8 51>95th 6 27 (1-4to23 8) (1-2to21-8)

Hyperactive behaviour (teacher)<95th 96 74 8-4 7-4>95th 4 26 (I1-6 to 42 0) (I1 4 to 36-4)

Hyperactive behaviour (parent)<95th 94 62 10-1 9 5>95th 6 38 (25 to403) (2-3to39-6)

ence in our results. In their study, reportedsnoring decreased with age probably because ofan increase in pharyngeal dimensions and adeno-tonsillar atrophy. They excluded children whoslept alone from their analysis because snoringwas reported considerably less often in thesechildren than in those who shared a room. Wedid not examine this factor in our study, thoughwe found that parents from lower socioeconomicgroups were more likely to have their assessmentof snoring confirmed during our overnightmonitoring, perhaps because of their morecramped living conditions. If we had alsoexcluded children sleeping alone then our pre-valence of reported snoring might have beeneven higher than 12 1%.Daytime sleepiness, restless sleep, and hyper-

active behaviour all showed a significant increas-ing trend across the snoring categories, hinting ata dose response relation. In adults, snoring isassociated with sleep disruption'9 and selfreported daytime sleepiness.20 Mild to moderatedaytime sleepiness is a subjective symptom andby relying on parental observation we are likelyto underestimate the relation. Hyperactivebehaviour has been reported in children withsleep and breathing disorders and largelyresolves after an adenotonsillectomy.4 7Although hyperactivity was subjectively definedin our questionnaire, its strong relation withsnoring in this large community sample, con-firming other clinical studies, suggests a trueassociation.We found no difference in the rates of reported

or video confirmed snoring between the sexes,similar to the findings of Corbo et al.'8 This is instark contrast with adults, where there is a largemale preponderance, presumably because themechanisms of pharyngeal narrowing in childrenare different from those in adults where neckobesity is a major factor.)

SLEEP STUDY AND BEHAVIOURAL ASSESSMENTSelection of children for overnight studyWe studied only 132/782 (16-9%) of the childrenfor whom we had completed questionnaires, andmay have therefore underestimated the trueprevalence of sleep and breathing disordersbecause of under reporting by parents. To studyall the children in the way we did would havebeen beyond our capacity. The alternativeapproach of studying a larger number of childrenwith simpler techniques, such as oximetry alone,would be inappropriate because upper airwayobstruction can cause significant sleep disturb-ance and daytime symptoms without hypo-xaemia7 21 and would furthermore not haveallowed us to determine the cause of anyoximetric abnormalities.Only one snorer was detected in the control

group compared with 15 snorers and seven withsleep and breathing disorders in the high riskgroup, indicating that our ranking of the child-ren according to the questionnaire responses hada physiological basis.

Comparison ofhigh risk and control groupsThe significant differences in sleep monitoring

364

on January 22, 2021 by guest. Protected by copyright.

http://adc.bmj.com

/A

rch Dis C

hild: first published as 10.1136/adc.68.3.360 on 1 March 1993. D

ownloaded from

Snoring, sleep disturbance, and behaviour in 4-5yearolds

results between the two questionnaire definedgroups indicates that parental responses reflectreal and measurable differences in their child-ren's sleep. Nevertheless only one third (22/66)of children in the high risk group actually snoredon the night of the study. Croft et al also foundparental history poorly predictive of whether a

child snored or not during a single night ofmonitoring.23

Are parents simply unreliable witnesses or dotheir responses represent an average statementabout their children's sleep stretching back overa considerable period of time? If the latter is truethen one night of sleep monitoring may notprovide an adequate description of a child'saverage sleep. It is possible that children fromthe high risk group have nasal and pharyngealdimensions that are critically balanced such thatminor upper respiratory infections could lead toupper airway obstruction, snoring, and restlesssleep. One of the questions used in ranking thechildren was the frequency of 'coughs and colds'(question 5, see appendix), and therefore child-ren in the high risk group had more upperrespiratory infections than the controls. Ourdecision not to study children with upperrespiratory infections until they had recovered islikely to have underestimated the problems thatchildren in the high risk group experience atnight. This may have biased our study againstconfirming the parental reports of snoring.Despite this possible bias we nevertheless foundsignificant differences in the overnight monitor-ing of the two groups. The relative weight thatshould be given to parental reports of a childs'sleep versus a night ofmonitoring is important indealing clinically with individual children and ininterpreting the results of large scale question-naire surveys such as this. No studies have yetaddressed the issue of night to night or evenweek to week variability in the results of sleepmonitoring in children with mild sleep andbreathing disorders. If night to night variabilityis the explanation for the negative sleep studies insome of our children from the high risk group(with their behaviour differences), then it wouldbe important to measure this variability asclinical decisions are often based on a single nightof monitoring. In severe cases one night isprobably sufficient.

Children from the high risk group were more

likely to come from a manual socioeconomicgroup, though after allowance for the effects ofparental smoking it was no longer significant.Maternal smoking exerted a more powerfuleffect than paternal smoking, presumablybecause this reflects more closely the child'spassive smoking environment. These findingsconfirm previous reports which have linkedpassive smoking, adenotonsillectomy,24 andsnoring in children.'8We shortened the Conners rating scale by

using only those questions which make upthree of the five subscales normally used. Thiswas to concentrate on those particular behaviourpatterns reported to be associated with sleep andbreathing disorders in children. The Connersscale is the most widely used research scale inchild psychopathology, with good stability over

time and inter-rater agreement.25 The scale is well

validated with significant agreement between theparticular subscales and direct observation ofchildren.26There were clear differences using this

behaviour rating scale between the two groups,high risk and control, identified by the originalquestionnaire. This strongly suggests thatpoorer behaviour in children is associated withdisturbed sleep due to upper airway obstruction.There was a considerable range in the Conners

scores of our children, similar to that found inother studies.27 The high risk group was concen-trated mainly at the higher end of this range, butonly about 10% of their scores actually exceededthe mean levels found in children referred to asouth London psychiatric clinic.27 Thus they areclearly not as disturbed as this referred group.

It is possible that our results are biased by atendency for some parents to generally be overcritical of their child, thus producing high scoreson the original questionnaire and the moreformal behavioural assessment. We think this isunlikely for two reasons, however. First, theteachers' behavioural assessments were alsodifferent for the two groups, albeit with lowerscores than the parents (a phenomenon that hasbeen noted before28). Second, there weremeasurable physiological differences duringsleep between the two groups showing that theanswers to the original questionnaire were notsimply reflecting different parental attitudes totheir children.

Other potential confounding factors such associal class and gender were allowed for in theanalysis and did not alter the results. Onepossible confounding variable that we have notformally examined is the known relationbetween behaviour problems and the deafness ofsecretory otitis media,28 which is itself associatedwith adenotonsillar hypertrophy. This isunlikely to be the explanation for our results foralthough teachers report more behaviour prob-lems in children with secretory otitis media, nosuch differences were found for parental reportsin the same study.28 Furthermore only twochildren in the high risk group (and none of thecontrols) had had hearing problems (due tosecretory otitis media, one with grommets). Twoclinical studies have found an associationbetween daytime sleepiness, behaviour prob-lems, and sleep disturbance due to snoring alone(that is, without hypoxaemia or sleep apnoea).2'These two studies showed improvements follow-ing adenotonsillectomy, supporting the viewthat this is a causal relation. In adults, snoring(without apnoea, hypopnoea, or hypoxaemia)can also cause sleep disruption,'9 and is associ-ated with daytime sleepiness.20 It is likely thatsimilar mechanisms operate in children, thoughthey may show other behavioural problemsbesides sleepiness. Many of the children fromthe high risk group did not have a significantsleep and breathing disorder on the night ofstudy and yet they displayed considerablebehavioural differences from the controls. Thissuggests either that the small but significantdifferences in nocturnal movement, desatura-tion, and pulse rate between the groups reflectbiologically important differences in sleepquality, or (if the parental history is to be

365

on January 22, 2021 by guest. Protected by copyright.

http://adc.bmj.com

/A

rch Dis C

hild: first published as 10.1136/adc.68.3.360 on 1 March 1993. D

ownloaded from

366 Ali, Pitson, Stradling

believed) that even intermittent snoring andsleep disruption can have deleterious effects onbehaviour.

CONCLUSIONSWe did not identify any children with severesleep and breathing disorders in this surveysuggesting that parents, general practitioners,and ear, nose, and throat surgeons identify andtreat most children with severe problems. Of thechildren with confirmed snoring or sleep andbreathing disorders that we identified, however,more than a third had a history of enlargedtonsils. Presumably in the past, when a moreliberal attitude towards tonsillectomy prevailed,these children and perhaps many of the otherchildren in the high risk group would have hadan operation. Children with a history of snoringare more likely to be sleepy during the day and tohave a number of 'difficult' behaviour patternsthan those without such a history. Longitudinalstudies, such as the Dunedin survey of otitismedia with effusion,28 are required to assess theimpact of these nocturnal and behavioural differ-ences on childrens' educational and socialdevelopment. It will then be possible to deter-mine if current criteria for adenotonsillectomyexclude children who might benefit from havingtheir tonsils removed. Such a study would alsohelp to define what constitutes a 'significant'sleep and breathing disorder in children.

N J Ali is a British Lung Foundation Research Fellow, D J Pitsonis supported by Wellcome Trust, and J R Stradling is a Wellcomesenior research fellow.

1 Brodsky L, Moore L, Stanievich JF. A comparison of tonsillarsize and oropharyngeal dimensions in children with obstruc-tive adenotonsillar hypertrophy. Int J Pediatr Otorhino-laryngol 1987; 13: 149-56.

2 Fry J. Are all 'T's and A's' really necessary? BMJ 1957; i:124-9.

3 Black N. Geographical variation in use of surgery for glue ear.J R Soc Med 1985; 78: 631-7.

4 Guilleminault C, Korobkin R, Winkle R. A review of 50children with obstructive sleep apnea syndrome. Lung 1981;159: 275-87.

5 Check WA. Does drop in T and A's pose new issue ofadenotonsillar hypertrophy?JAMA 1982; 247: 1229-30.

6 Rosenfeld RM, Green RP. Tonsillectomy and adenoidectomy:changing trends. Ann Otol Rhinol Laryngol 1990; 99: 187-91.

7 Stradling JR, Thomas G, Warley ARH, Williams P, FreelandA. Effect of adenotonsillectomy on nocturnal hypoxaemia,sleep disturbance, and symptoms in snoring children. Lancet1990; 335: 249-53.

8 Office of Population and Census and Surveys. Classification ofoccupations. London: HMSO, 1980.

9 Trites RL, Blouin AGA, Laprade K. Factor analysis of theConners teacher rating scale based on a large normativesample. J Consult Clin Psychol 1982; 50: 615-23.

10 Vos PJE, Stradling JR. Assessment of sleep times andmovement arousals from video recordings. J3ournal ofAmbulatory Monitoring 1991; 4: 35-42.

11 Warley AR, Mitchell JH, Stradling JR.- Evaluation of theOhmeda 3700 pulse oximeter. Thorax 1987; 42: 892-6.

12 Stradling JR, Crosby JH. Relation between systemic hyper-tension and sleep hypoxaemia or snoring: analysis in 748men drawn from general practice. BMJ 1990; 300: 75-8.

13 Conners CK. A teacher rating scale for use in drug studies withchildren. AmJr Psychol 1969; 126: 884-8.

14 Cody RP, Smith JK. Applied statistics and the SAS programminglanguage. New York: Elsevier, 1987.

15 Payne C. The GLIM systems release 3.77 manual. Oxford:Numerical Algorithms Group, Royal Statistical Society,1985.

16 Altman DG. Practical statistics for medical research. London:Chapman and Hall, 1991.

17 Gardner MJ, Altman DG. Statistics with confidence. London:BMA, 1989.

18 Corbo GM, Fuciarelli F, Foresi A, De-Benedetto F. Snoringin children: association with respiratory symptoms andpassive smoking [published erratum appears in BMJ 1990;300: 226]. BMJ 1989; 299: 1491-4.

19 Guilleminault C, Stoohs R, Duncan S. Snoring (I). Daytimesleepiness in regular heavy snorers. Chest 1991; 99: 40-8.

20 Stradling JR, Crosby JH, Payne CD. Self-reported snoringand daytime sleepiness in men aged 35-65 years. Thorax1991;46: 807-10.

21 Guilleminault C, Winkle R, Korobkin R, Simmons B.Children and nocturnal snoring: evaluation of the effects ofsleep related respiratory resistive load and daytime function-ing. Eur3r Pediatr 1982; 139: 165-71.

22 Davies RJ, Stradling JR. The relationship between neckcircumference, radiographic pharyngeal anatomy, and theobstructive sleep apnoea syndrome. Eur RespirJ 1990; 3:509-14.

23 Croft CB, Brockbank MJ, Wright A, Swanston AR. Obstruc-tive sleep apnoea in children undergoing routine tonsillec-tomy and adenoidectomy. Clin Otolaryngol 1990; 15: 307-14.

24 Said G, Zalokar J, Lellouch J, Patois E. Parental smokingrelated to adenoidectomy and tonsillectomy in children.J Epidemiol Community Health 1978; 32: 97-101.

25 Trites RL, Blouin AGA, Ferguson HB, Lynch GW. TheConners teacher rating scale: an epidemiological inter-raterreliability and follow up investigation. In: Gadow K, LoneyJ, eds. Psychosocial aspects ofdrug treatment for hyperactivity.Boulder, Colorado: Westview Press, 1981.

26 Rapoport JL, Benoit M. The relation of direct home observa-tions to the clinic evaluation of hyperactive school age boys.J Child Psychol Psychiatry 1975; 16: 141-7.

27 Taylor E, Sandberg S. Hyperactive behaviour in Englishschoolchildren: a questionnaire survey. J Abnorm ChildPsychol 1984; 12: 143-56.

28 Chalmers DC, Stewart I, Silva P, Mulvena A. Otitis mediawith effusion in children. Clin Dev Med 1989; 108.

AppendixQuestionnaire used in the postal survey.1. Doesyour child have a problem related to sleeping?1=Never 2=Rarely 3=Sometimes 4=Often2. Does helshe have trouble getting to or staying asleep?1=Never 2=Rarely 3=Sometimes 4=Often3. Does helshe fall asleep during the day particularly whennot active (for example watching TV or reading)?1=Never 2=Rarely 3=Sometimes 4=Often4. Is helshe considered to be hyperactive?1 =Never 2 =Rarely 3= Sometimes 4=Often5. Does helshe get coughs and colds?1=Never 2=Rarely 3=Sometimes 4=Often6. Does helshe snore at night?1 =Never 2=Rarely 3= Sometimes 4=Most nights

(even without a cold)7. Is the volume of the snoring?1=Quiet 2=Moderate 3=Loud 4=Very loud8. Does helshe sweat at night?1= Never 2= Rarely 3= Sometimes 4= Often9. Isyour child a restless sleeper?1=Never 2=Rarely 3=Sometimes 4=Often10. Doesyour child sleep with hislher head tipped back?1=Never 2=Rarely 3=Sometimes 4=Often11. Does your child tend to breathe through hislher mouthrather than nose?1=Never 2=Rarely 3=Sometimes 4=Often12. Have you ever been told that your child has enlargedtonsils?l=No 2=Yes13. Hasyour child had hislher tonsils taken out?l=No 2=Yes

on January 22, 2021 by guest. Protected by copyright.

http://adc.bmj.com

/A

rch Dis C

hild: first published as 10.1136/adc.68.3.360 on 1 March 1993. D

ownloaded from