Criteria for the Detection of Sleep-Associated Bruxismin Humans

Takashi Ikeda, DDS, PhDAssistant ProfessorDepartment of Fixed Prosthodontics

Keisuke Nishigawa, DDS, PhDAssistant ProfessorDepartment of Fixed Prosthodonlics

Tokushims UniversitySchooi of DentistryTokusliima, Japan

KaruoKondo. DDS. PhDPrivate PracticeEhime, Japanand Visiting Scholar (1969-1991)University of Caiifornia, Los AngelesLos Angeies, Caiifornia

Hisahiro Takeuchi. DDS, PhDAssistant ProfessorDepartment of Fixed ProsthodonticsTokushima UniversitySchooi of DentistryTokushima, Japan

Glenn T, Clark, DDS, MSAssociate Dean of Research,

Professor, and ChairUniversity of California, Los AngelesSchool of Dentistry, Diagnostic

Sciences and Crofacial PainLos Angeles, Caiifornia

Correspondence to:Dr Gienn T, Cisrk43-009A Center for Health Science10833 Le Conte AvenueLos Angeles, California 90095-1668

Surface electromyagraphy of the masseler and electrocardiogramrecordings of heart activity during sleep were performed on ninesubjects who suffer from an oral motor dysfunction (bruxism)during sleep. Signals tvere monitored in the subject's home sleep-ing environment over 4 consecutive nights. A total of 36 nights ofdata were analyzed to perform the following: (1) describe thenature and magnitude of total masseter muscle eiectromyographicactivity above a minimum threshold of 3% of each subject's indi-vidually established maximum voluntary contraction level; and ¡2)describe electrocardiograph rate changes (using the R-R interval)that occurred in relation to these eiectromyographic elevations.From these data, criteria for detection of bruxism events wereestablished and combined into a fully automated event detectionalgorithm. The mean number and duration of the detected brux-ism events are reported. The underlying logic for tbe criteriaselected, and what effect other possible criteria would have on theseparation of abnormal from normal motor events, is also pre-sented and discussed.J OROFACIAL PAIN 1996;10:27O-282,

key words: bruxism, detection criteria, sleep, masseter muscle,electromyography

In the past, sleep-associated bruxistn was studied withpolysomnographic methods (including surface electromyogra-phy tracing from the jaw elevator muscles). The presence or

absence of hruxism and its frequency were determined frotn theserecordings using a combination of the atnplitude and pattern ofthe eiectromyographic (EMG) tracing recorded from the masseteror tempotalis muscle. Over the years, such polysomnographicrecordings have adequately described the general characteristicsand nature of bruxism. Unfortunately, however, these visualassessment methods do not lend themselves to reliable quantifica-tion of the behavior.

What is known about bruxism currently is that the jaw elevatormuscles commonly exhibit two basic patterns of activity: rhyth-mic, chcwinglike movements; and prolonged, strong isotonic con-tractions,' The total duration of btuxism per night is approxi-mately 10 minutes in patients with clinically obvious signs ofbruxism {eg, attrition).^'^ It has also been reported that most brux-ism episodes appear during stage I and stage 2 sleep,''" In addi-tion, bruxism occurs frequently during partial sleep arousal (whensleep moves from a deeper to a lighter stage).^ Some bruxismepisodes have been reported to be accompanied with mcreasedheatt and respiratory rates, ••^•^

270 Volume 10, Number 3, 1996

Ikeda et al

Bruxism has been described in the literaturesitice the turn of the cetitury. Until the arrival ofmtiiticbiinnel sleep laborarory studies in tbe 1950s,however, the actual behavior of brnxistn had notbeen measured. Iti tbe 1970s, portable EMC re-cording devices were developed by Riigb and col-leagues - to record masseter muscle activity in tbesubject's botne environment. Rtigb's device pro-vided cumulative totals of nightly electrical activityabove 20 pV and allowed new and itnportant dataabout tbe levels of masseter muscle activity to begatbered in tbe patient's usual sleeping environ-ment. iVlost of the studies using Rugb's portableEMG device used a 20-pV thresbold to preventrecording of minor masseter contractions pre-sumed to be not significant in clencbing and grind-ing. Clark and colleagues'""'^ utilized Rugh'sportable EMG recorder in various studies tbatrecorded the bruxism levels in patients and evalu-ated the effect of an audio alarm method for sup-pression of the bruxism response. In general, tbesesingle-channel EMG-based home measurementsystems bave helped acbieve a greater understand-ing of bruxism. Tbe disadvantages of tbis methodare that second-by-second levels of bruxism arenot available for study, and correlation betweenmotor activity and sleep stage level is not possible.In 1981, Wagner "" developed anotber single-chan-nel EMG home-based recording system. Wagner'ssystem involved an EMG amplifier, filter, and inte-grator; unlike Rugb's portable cumulative EMGrecording system, it defined bruxism as activitygreater than 5 pV (integral average). Wagner alsodetermined tbar 3 seconds of quiescent EMGrecording was required to tally separate bruxingepisodes. Summary scores of bruxism levels pernight were available from Wagner's data.

in 19S2, Piccione and colleagues'^ defined brux-ing events using masseter EMG amplitude, rhyth-miciry, and duration criteria from a polygraphicrecording. They proposed tbat a bruxism eventwas present only if the EMG amplitude exceeded20 (iV. Further, the EMG bad to rise above tbislevel for no less than 0.5 seconds and no longerthan 1.5 seconds to be a bruxism event. They alsorequired that bruxism events occur in a series oftwo or more rbytbmic contractions separated byno more tban 2.5 seconds. Piccione and colleaguesdid not score any prolonged EMG elevations (>1.5 seconds) nor were solitary elevations scored asbruvism. In 1983, Stock and Clarke'^ described abruxism detection system tbat consisted of a bat-tery-operated analog EMG device, an optoisolater,an analog-to-digital (A/D) converter, a micropro-cessor, and a cassette tape recorder. Bruxism

events were selected automatically by tbe device ifthey satisfied three criteria: (1) EMG elevationshad to be more tban tbe fourtb least significant bitof an S-bit amplifier whose gain was adjusted tonot trigger with ordinary orofacial movements^ (2)duration of EMG elevations should be greater thanor equal to 2 seconds; and (3) gaps of 1 second orless between the elevations were joined and consid-ered a single event.

In 1988, using yet another set of criteria. Wareand Rugb- compared tbe amount and topographyof bruxing events m various sleep stages during 1night of sleep across three groups of patients. Anepisode was scored as bruxism if it met the follow-ing four criteria: (1) the EMG signal was seen asan artifact on the electroencephalogram (EEG)tracing; (2) it was greater than 75 pV or it wastwice the background activity; (3) the phasic eventswere repetitive with two or more discrete bursts;and (4) the phasic events lasted more than 0.25seconds. During tbe scoring, bruxism duration wasmeasured, and the number of discrete (phasic)bursts were counted. In 1990, Okeson et aP"defined and detected a bruxing event if activit}' ofthe masseter muscle exceeded 40% of its maximumclench and lasted 2 or more seconds. In 1992, VellyMiguel et al ' adopted these same criteria.

Although these previous studies have allowedrhe development of better ideas about the natureof bruxism, it is clear that quite different criteriaare used to measure tbe level of bruxism (Table 1).If research about the nature of bruxism is to con-tinue, a standard set of criteria for wbat consti-tutes a bruxism event needs to be establisbed. Inaddition, regardless of wbether these recordingsare performed in a sleep laboratory or in tbe sub-ject's home environment, most researchers suggestthat tbe recording technique involve tbe collectionof several consecutive nights of data because of itsvariable nature. Because the technology is nowwithin reach, the recorded data should be scoredusing a semiautomated or fully automated com-puter method to reduce measurement error andspeed up tbe analysis. In tbe past, the polygraphicrecords of bruxism have been scored by an expertusing visual analysis metbodology, Altbough it issomewhat tedious to score a typical 7-bour sleeptracing by hand, it requires no sopbisticated equip-ment, and artifacts are easy to detect. The disad-vantage is that such an interpretation requires asubstantial amount of expertise and is often notrepeatable between different individuals scoringtbe same record.

The advantages of using a semiautomated orfully automated computer scoring routine are in-

Journal of Orofaeial Pain 2 7 1

ikeda et ai

Table 1 Review of Bruxism Event Detection Criteria (1975-1992)

RughandSolber9=C197S)

Wagner'" (19811

Piccioneet ai'^n982)

Stock and Ciariie's 11933)

Kydd and Daly^ (1985)

Ware and RiJgb^ (1988)

OkesQretai"(1990)

Veily Miguel et ai'=11992)

Amplitude

>20mV(intégrai average)>5mV(intégrai average)>20mV(intégrai average)A ievei greaterthan that producedby ordinary orofaoialmovementNone

Twice the backgroundor > 75 mV (raw)> 40% MVCpeak (raw)> 40% MVCpeak (raw)

Duration

None

None

>0,5s< 1.5 s> 2 s

None

> Q.25 s

> 2 s

> 0.25 s

Linkage

None

< 3 s

<2,5s

< t s

None

None

None

None

Other crireria

None stated

None stated

None stated

None stated

Movement artifactsand eiectricsinoise were deietedNone stated

None stated

None stated

Analyzing method

Analog device

Analog deviceand msnuaiManuai

Anaiog deviceand computeraigorithm

Manuai

Manusi

Manuai

Manual

num voiunlary corilrgcdon.

creased reliability and reproducibility. In addition,if the computer algorithm is powerful enough toincorporate one or more channels of data besidesthe masseter EMG channel, it is possible to betterdefine the hruxism event or assess its correlationwith other physiologic events (eg, the extent that abruxism event is associated witb a sleep statechange).

The present study had three purposes: (1)describe the EMG and electrocardiographic (EGG]data that were recorded from a group of bruxismsubjects; (2) using this data, select the most logicalcriteria to be used in a computer algorithm todetect bruxism events; and (.3) describe in detailthis algorithm and present detailed descriptive dataon the level of bruxism observed in tbe subjects.

Materials and Methods

Subjects

Individuals witb self-acknowledged facial pain andclinically confirmed dental attrition, who re-sponded to an advertisement in the campus news-paper and satisfied the inclusion and exclusion cri-teria (see below), were selected as subjects.Potential subjects bad to meet the following inclu-sion criteria: (1) be between 18 and 65 years ofage; (2) exbibit evidence of tooth attrition suchchat at least one tooth (usually canine) had dentine

exposure on the occlusal or nicisal surface; (.3) hein good health; (4) have tried in the past, but notpresently be actively using, an occlusal coverageacrylic resin appliance; and (5) have stated theyhave frequent jaw or tooth pain in the morningand have demonstrated moderate to severe tender-ness in two or more sites in their jaw closing mus-cles (masseter and temporalis). The tenderness rat-ing was elicited via the application of 1.8 kg ofpressure to the muscle for 2 seconds with a hand-held pressure algometer device (Pain Diagnosticsand Thermography, White Plains, NY). Afterexperiencing the pressure, the subjects selected anumber from 0 to 3 tbat represented their painlevel (0 = none, 1 ^ mild, 2 = moderate, 3 -severe). Exclusion criteria for the subjects were thefollowing: (1) the daily use of any medications oralcohol; (2> the inability to sleep with a recordingapparatus attached; and (3) the need for anyurgent dental treatment.

Study Design

The data collection setup used in this descriptivesrudy is shown in Fig 1. Masseter muscle EMGactivity was measured using miniature bipolar sur-face electrodes, Electrocardiographic activity wasrecorded by placing electrodes over tbe subject'sleft clavicle, sternum, and left lateral chest wall.The signals from tbe various electrodes wereamplified and recorded onto a portable multichan-

272 Voiume 10, Number 3, 1996

Ikeda et al

Recording

Analyzing

Cassetterecorder

(TEAC MR-10)

Eight times speed

Amolifier

EMG 1 1 ^ ^

ECG 1 " ^ ^

RESP 1 1 ' ^

Condtioning amputier

- J E M ^ ^ RMS 1 -

Window 1discriminator 1

10 ms 5 V pulse

—flESP'^r-Band | _

pass tiiter 1

0 to 10 Hz

t

fc

fc

Portablecassetterecorder

(TEAC HR-IOJ)

A/D

250 Hz

2,000Hz

31 Hz

Computer

Monitor

Printer

Samplingrate

Fig 1 Respiration (RESP), ECG, and EMG data were collected by a portable cassette recorder |TEAC HR-IOJ) in thepatients" homes. These data were then reproduced by a cassette recorder (TEAC MR-IO| at a speed eight times fasterthan the original; then they were transformed, acquired to 8-bit digital data, and stored In a computer.

nd analog tape recorder (TEAC HR-IOJ, cassettedata recordet). All recordings wete performed inthe subject's home for 4 consecutive nights, unlessa technical problem occurred (eg, electrode detach-ment). In the latter case, an additional recordingnight was performed as soon as the problem wasdiscovered.

Data Reduction and Acquisition to Computer

The analog-tape-recorded data were first condi-tioned and then acqtiired to computer in a digitalformat. This process involved playing back thedata on a second device (TEAC, MR-10) at a tapespeed that was eight times faster than the originalrecording speed. Next, using a signal conditioningdevice (SA-414 Analog Processor, SA Instrumen-tation Service Associates, Encinitas, CA), the rawEMG signal was amplified and converted to a

smooth EMG signal using a root mean square(RMS) conversion. The electrocardiographic(ECG) signal was conditioned by first detecting theR wave of the ECG tracing using a wmdow dis-criminator. Each detected R wave then triggered a10-ms 5-V square wave pulse. These pulse andsmooth EMG data were then transferred by an 8-bit A/D converter and data acquisition system tothe computer. The signal sampling rate of theRMS-converted EMG signal was 250 Hz, and theR wave triggered pulse signal was sampled at2,000 Hz. Respiration data was filtered with aband pass filter between 0 to 10 Hz and sampledat 31 Hi. Eor final analysis and display of the data,the amplitude of the EMC signal was quantified interms of the subjects' 100% maximum voluntarycontraction (MVC) level. The 100% MVC foreach subject was recorded at the beginning of eachnight using three brief (2-second) MVC efforts.

Journal of Orofacial Pain 2 7 3

ikeda et ai

Data Analysis

The acquired data WLTC analyzed tn two stages byusing a seminiitomated or fully automated custotncomputer program. First., all EMG elevations abovea 3 % MVC were qtuintified with regard to the dura-tion, peak atnplitudif, and number of elevations perhour for the entire data set of 36 nights. The R-Rwave interval data during these 3% MVC or greaterEMG elevations were converted to beats-per-rnmutedata and also quantified. The second stage of theanalysis involved estahlishing and testing two addi-tional thresholds (10% atid 20% MVC) and threeother criteria: (I) the minimum time needed betweenEMG elevations before they were cotisidered sepa-rate events; (2) the minimum EMG duration neededto he considered an event; and (3) a minimum beats-per-mmute (bpm) täte increase occurrmg in conjitnc-tion wtth an identified EMG elevation needed to bean event. This last criterion used beats-per-minutedata during the RMC elevation and during a non-EMG elevation period occurring immediately beforethe event. The null hypothesis that the number ofevents per hour did not change across the 4 recordeddays was also tested. Eor this analysis, two of thenine subjects were excluded because they had torerecord a night of data because of equipment mal-function. Only data that were collected across 4 con-secutive nights were included. A regression line wasthen fit to each subject's nightly data ¡events perhour). The weighted mean slope of these lines wereestablished for each of the three different thresholdcriteria (3%, 10%, 20% MVG). Weighting of rhesub|ects' data was performed because some subjectshad better fitting linear trends for their data thanothers. To incorporate this feature., an inverse regres-sion coefficient variance estimate was used to weightthe linear trends. The weighted mean slope lineswere then tested usmg a one-sample two-trial t testto see if they were significantly different from zero.Finally, the resulting bruxism event data were bothdescribed and displayed using two- and three-axisplots. The underlying logic for the above criteria andthe results of the above analysis are presented in thenext section. A hrief abstract of these resuirs hasappeared previously.'^

Results

Subject Demographics

Nine subjects (five men and four women] com-pleted the protocol. Their mean age (± one stan-dard deviation [SD]) was 26.2 + 4.18 years.

Sleep Data Description

The subjects in this experiment had a tota! recordingtime of 208 hours. The mean recording time (± oneSD) per subject was 5.78 ± 0.53 hours per night.

Electromyographic Data Levels Greater Than3% MVC. When a threshold of 3% EMG wasused, there were 3,339 EMG events within the 36nights of recorded data. Because a 3% MVC eleva-tion would not be considered a strong contracrionof the jaw elevators, additional higher amplitudethreshold criteria were also considered. When thethreshold criteria were increased to 10% and 20%MVC, the total number of qualifying eventsdecreased to 2,601 and 1,706, respectively.

Electromyographic Linking Data. Bruxism willsometimes occur as a closely linked series of con-tractions. Bruxism is often associated with a sleepstate change and a generalized cardiorespiratorysystem arousal. Eor these reasons, it was decidedto iink EMG elevations that were temporallyrelated. The rationale for linking closely related£MG élévations was that this would allow the car-diovascular state changes during an episode ofbruxism (whether it occurred as a single contrac-tion or as a ltnked series) to be better determinedand compared to the cardiovascular status duringa preceding non-EMG elevation period. A mini-mum linkage time of less than or equal to 5 sec-onds between EMG offset and next onset wasselected, and all qualifying EMG elevations thatsatisfied these threshold and linkage criteria werenow considered as a single potential bruxismevent. As a result of adding this criterion, therewere now determined to be 2,078 potential brux-ism events (using 3% MVG) within the 36 nightsof recorded data. When the threshold criteria wereincreased to 10% and 20% MVC, the total num-ber of qualifying potential bruxism events afterlinkage decreased to 1,570 and 1,057, respectively.

Electromyographic Duration Data. There arehigh amplitude EMG elevations of short durationthat occur as myoclonic twitches or as a result ofBMG signal or line artifacts. Therefore, all EMGelevations tbat were above threshold for less than3 seconds from onset to offset were eliminated.When a minitnum EMG duration of 3 seconds wasused in cotnbination with the 3% MVC and the 5-second linking criteria, there were 1,538 potentialbruxism events within the 36 nights of recordeddata. When the threshold criteria were increased to10% and 20% MVC, the 3-sccond minimumduration criteria reduced the total number ofpotential bruxism events after linkage decreased to920 and 526, respectively.

274 Voiume 10. Number 3. 1996

Ikeda et al

300

250

200

1 150od

100

50

0 • - . • . i l l-30 -25-20-15-10 -5 0 S

n = 1,538

1II.Mil

10 15 20 25 30 35 40 45 50 55 60 65 70

ECG cliang ng (%)

Fig 2 Histogram showing the heart rate changes that occurred in association with 1,538EMG elevations. These 1,538 elevations satisfied ali three minimum EMG criteria (3%MVC threshold, linkage, and duration).

Electrocardiographic Rate Change Data. Asmentioned above, bruxism is a phenomenoti thatTypically occurs in association with a generalizedcardiorespiratoty system arousal. For this reason,the fitial event detection criterion was a minimumbe at s-per-minute rate change of greater than 5%.This magnitude of change would equate to anincrease from 60 bpm to 63 bpm within a 5-sec-ond period. The comparison period for this assess-ment was the 5-second time period immediatelypreceding the detected potential bruxism event,"When a greater than 5% bpm rate change increasecriterion was added to the previously discussedEMG criteria (linkage and duration), there weredetermined to be 1,270 bruxism events within the36 nights of recorded data for the 3% MVCthreshold. When the EMG threshold criteria wereincreased to 107o and 20% MVC, the total num-her of hruxism events decreased to 756 and 414,respectively.

Figure 2 shows a histogram of the number oftimes a heart rate change within the specifiedtange occurred in association with the detectedbruxism events (n = 1,538) that satisfied all threeminimum EMG criteria (3% MVC threshold, hnk-age, and duration). The most common heart ratechange was a 10% to 14.9% increase. In this fig-

ure, 82,6% of the potential bruxism events alsohad a greater than or equal to 5% increase in bpm.The change in EMG amplitude threshold to 10%and 20% MVC did not substantially altet this pat-tern. Fot a 10% MVC EMG threshold, 82,2% ofthe potential bruxism events wete above the 5%bpm increase; for the 20% EMG amplitude thresh-old, this figure was 78,8%.

Peak HMG Elevations and Duration of BruxismEvents. A histogtam of the 1,270 bruxism eventsthar were above the 3% MVG threshold and thatsatisfied the other described criteria (linkage, dura-tion, and ECG change) is presented in Fig 3. Eachbar in this figure represents the number of bruxismevents that had a peak amplitude in the specifiedrange. From this figure, it can be seen that themost common bruxism events had a peak EMGelevation between 10% and 14,9% of MVG.

A histogram of the 1,270 bruxism events with aspecified duration are presented in Fig 4. As in Fig3, these bruxism events wete those that weteabove the 3% MVC threshold level and satisfiedthe described minimum linkage, duration, andECG change criteria. Each bat in the figure repre-sents the number of bruxism events that had adutation in the specified rime tange. From this fig-ure, it can be seen that the most common duration

Joumai of Orofacol Pam 275

Ikeda et al

180

ISO

140

a 120

- 100

No.

60

40

20

-

-

-

-

-

-

1

n = 1,270

iHili

l l l . a

Illllll10 20 30 40 50 SO 70 80 90 100 110 120 130 140 150 180 170 180 190 200 210 220

% MVC

Fig 3 Histogram showing the 1,270 EMG elevations that were above the 3% MVC threshold. These 1,270 EMG ele-vadoiis also satisfied all three criteria (linkage, duration, and ECG rate change).

3 7 11 15 19 23 27 31 35 39 43 47 51 55 59 63 67 71

Duration (seconds)

Fig 4 Histogram showing the duration of 1,270 tMG elevations that were above the 3% MVC threshold. The EMGelevations also satisfied all three criteria (linkage, duration, and ECG race change).

276 Volume 10, Number 3, 1996

Ikeda et ai

Table 2 I

EMGthreshold

> 3% MVC>10% WVC> 20% MVC

îruxism Events at Three Different EMG Thresholds

Mean EMG durationper cvent(s)

12,9 ± 1,910.1 ±2.18.1 ± 1.6

Mean bruxism eventsper hour (No.)

6.2 ± 1.63,6 t 1.22.0 ± 1.0

Mean peak EMClevel (% MVC)

36.1 ±10,646.3*10,356,5 ± 9,7

Table 3 Bruxism Events for Each Recording Day (n - 7)

Mean (± SD) events pei hour

3% MVC

Day 1 5.7 ±2,0

Day 2 5.7 ± 1.7Day3 6.1 ±2.5Day 4 6.0 ± 1.3

10% MVC3.0 ± 1,43,7 ± 1,34,5 ±2,1

20% MVC1 5 ± 0 72.1 ±0.92.6 ± 1.52,3 - 1.3

Regression Coefficient for Each Recording Day (n = 7)

3% MVC10% MVC20% MVC

Regression coefficient weight

-0,08 + 0.42-0,40 + 0.46-0,45 + 0,41

P values

,594,043.017

was hetween 3.0 and 4,9 seconds. Tbe maximumvalue in these data was 72.8 seconds.

The mean (* one SD} duration per bruxismevent, number of bruxism events per hour, andpeak EMG level per event tbat satisfied the threedifferent qualifying thresholds are presented inTable 2.

Order Effect Data. To see wbether the numberof events across each recording day varied in a sys-tematic fashion, the total number of bruxismevents (using the 3%, 10%, and 20% iMVC ampli-tude criteria) per day was determined and com-pated for significant differences. The mean number(± one SD) of bruxism events per bour, theweighted mean tegression coefficients fit to thesedata, and the P values from the í test are presentedin Table 3. The data revealed no statistically signif-icant differences for the 3% MVC criteria, but the10% and 20% MVC criteria events did show astatistically significant difference across the 4 daysof recording.

Final Bruxism Criteria Data. Eigure 5 showstwo bruxism events from a single subject who metthe four described criteria (EMG tbreshold = 10%MVC). Tbe area below tbe elevated EMG signalthat defines tbe event is marked with a solid dou-

ble line. In Eig 6, a single subject's entire night ofEMC data is displayed using a three-axis plot,

Einal Bruxism Event Detection Algorithm. Tbelogic by which the above algorithm works isdemonstrated in Eig 7. Each EMG potential eleva-tion needed ro satisfy four criteria (EMG thresh-old, linkage, duration, and ECG change) in asequential fashion to qualify as a bruxism event.Detection of the event was automatically per-formed by customizing a software program.

Discussion

Until an accurate and reasonable cost effectivemethod of measuring the occurrence of bruxism isavailable, no substantial progress will be made inits understanding and subsequent treatment. Themethodology proposed to analyze and quantifybruxism events could also be applied to study anyepisodic motor behavior (eg, periodic leg move-ment disorder). Of course, instrumented sleepstudies performed in a hospiral-based sleep disor-ders clinic can be used for the diagnosis of brux-ism,^'^-' Unfortunately, most patients are unableto record multiple nights without considerable

Journai of Orofacial Pain 277

Ikeda et al

I I I I I I I I I J I I III IECG

Threshold level for EMG

EMG

Respiration

Time 203 204

Fig 5 Two brtixisni events (2 minutes of actual recordingl. The small vertical lines at the top oftracing represent a heart beat. In the middle upper portion of the tracing, the continuons hori-zontal line represents the RMS-smooched EMG data. The horizontal straight line superimposedon the EMG data line is the 10% MVC level. When the EMG line is above this 10% level andsatisfies the previonsly described hnkage and duration criteria, a single horizontal line appearsbeneath the EMG elevation. Eor this EMG elevation to be qualified as a bruxism event, it mustalso satisfy ECG rate change criterion. All qualified hruxism events have a second horizontalline beneath the EMG elevation. The other continuous horizontal line with variable amplituderepresents the respiration changes. These data were not used for any analytic purpose in thisstudy.

expense. Eurthermore, the variable nature of brux-ism often results in an unproductive recording dur-ing the first 1 or 2 nights during a sleep laboratoryrecording session. Our data reconfirmed this find-ing by showing a small but consistent increase inthe level of hruxism across the 4 nights. The mag-nitude of this change might actually be smaller inour study than in prior reports because recordingswere performed in the subject's own sleeping envi-ronment. We speculate that a more dramaticchange in the sleep environment, as would occur ina hospital-based sleep laboratory, might have astronger influence on bruxism levels during thefirst night. Eor this reason, we recommend that aminimum of 4 nights of recording be utilized inbruxism studies.

In designing the event detection algorithm forbruxism, it was decided that using a combinationof EMG and heart rate changes might be helpful.This decision was based on reports that brtjxism

events are consistently accompanied by an increasein heart rate and respiration.^- ' The observationsof these prior authors were clearly confirmed byour data.

In selecting the EMG amplitude to use as a mini-mum threshold, prior research'^'"* suggested brux-ism event detection tbresholds as high as 40% ofMVC. These authors made this determinationbased on raw peak EMG and not smoothed inte-gral average EMG (as used in our study). They didnot provide any data showing that this was a logi-cal threshold to utilize, nor did they establish howmany of the recorded elevations in the populationwere above 40% versus below 40%. Based on thedistribution of RMS-converted peak EMG activityin our data, we elected to use 10% as the mini-mum threshold. A 20%, and most certainly a40%, MVC level would be too high to use as aminimum event detection threshold when usingsmoothed integral averaged EMG. It is certainly

278 Volume 10, Number 3. 1996

Ikeda et al

100

% MVC

oïruxism event

No. of events: 36Total time: 267.4 seconds

hour Total anaiyzed time;418 minutes

fig 6 Example of a whole night's EMG data from a single patient. The oblique axis lines represent the actual recordeddata. Kach oblique line, beginning ac tlic bottom left of the horizontal line, is .10 minutes of recording. The vertical axisis the level of EMG quantified in a percent of subject's maximum voluntary contraction level. Each short horizontal litierepresents turn markers repeating every 2 minutes. Along rhe bottom horizontal axis, each 30-mjhutc epoch of EMGbegins and then follows the oblique lines. A small horizontal dash can be seen next to oblique line; this dash indicatesthat the EMG elevation satisfied the algorithm and was classified as a bruxism event.

debatable regarding how harmful an isolated 10%MVG elevation would be; however, because thestated goal of the research is to record all substan-tial bruxism elevations, we believe the 10% crite-rion appears appropriate. By changing the thresh-old criteria from 3 % to 10% MVC, 22 .1% of thepotential bruxism events were eliminated. If 20%MVG threshold criterion was used, 48.9% of thepotential bruxism events would have been elimi-nated.

Approximately 3 8 % of the potential bruxismevents were affected by the linkage criterion. Thismeans that 62% of all EMG elevation were fartherapart than 5 seconds. A strong reason for selectionof a minimum linkage duration of 5 seconds wasthat to calctjiate any heart rate change, a minimumpreceding period of at least 5 seconds without anEMG elevation was needed to allow the ECG rateto return to a nonaroused level.

With regard to the 3-second duration criterionfor a bruxism event, only 25.9% of the 3 % MVC

events were eliminated^ however, 41 .4% and50.2% of 10% and 20% MVG events were elimi-nated, respectively. The selection of 3 seconds asthe minimum duration criterion was so that atleast two and hopefully three heart beats perminute were available to calculate the EGG rate. ItshoLild also be noted that selecting a high EMGthreshold decreases the duration of a bruxismevent if the tirne-above-threshold is used to mea-sure the length of the event. The mean duration ofEMG events ahove 10% MVG (after linking) was10.1 seconds. This %'alue was only slightly longerthan those reported by other researchers. Redingand colleagues*' reported 9 seconds as the typicalduration of what they described as bruxism rhyth-mic contractions. Glarke et a l " reported bruxismevents to be 7.8 seconds. Unfortunately, priorreports on duration of bruxism events do notclearly specify how the duration was calculated.We believe the selection of a 3-second minimumduration appears appropriate for our algorithm

Journal of Orofaciai Pam 2 7 9

Ikeda et al

Bruxjsm event

Not selected

Two events combined I

(yes)Criterion 1 (EMG threshold)> 10% ol the subject's MVC

Criterion 2 {EMG event interval)< 5 seconds between events

(offset to next onset)

Fig 7 Bruxism eventdetection algorithm.

because most events will be detected and most arti-facts (eg, myoclonic twitches and EMG artifacts)will be rejected.

Okeson and colleagues^" reported that heart rateduring bruxing events increased on average by16,6% with a range from 6,1% to 40,2%, butthey did not suggest a minimum EGG rate changethat should serve as a criterion for detection. In

our data, we also report an average ECG ratechange to be between 10% and 20%, and the vastmajority of these ECG rate changes associatedwith EMG elevations were clearly mote than 5%.For this reason, the 5% EGG criterion was reason-able to include. The addition of this ctitetioncaused an additional 18% of the EMG elevation tobe eliminated.

280 Volume 10, Number 3, 1996

Ikeda et al

Overall, tbe number and duration of bruxismevents reported in tbis study using our four-crite-rion computer algorithm corresponds well to theprior literature where traditional polysomno-grapbic recordings were tediously scored by band.Tbe mean number of bruxism events reported inthis study using tbe 10% MVC threshold criterionranged between 3.1 and 4,8 events per hour. Redingand colleagues'* reported 4.4 events per bour, andClarke and colleagues" indicated 5 events per hourfor bruxism patients. Considering all tbe data, weare confident that this algorithm does not overre-port or underreport bruxism events. A visual as-sessment of tbe tiiree-axis diagram of nightly EMGelevations shows that a vast majority of all sub-stantial EMG elevations were detected as bruxismevents by the algoritbm.

In summary, the event detection algoritbm pro-posed and examined in this study has been logi-cally determined and is now awaiting additionalvalidation by other laboratories. Witbout question,the use of defined criteria and a semiautomated orfully automated computer-based algorithm allowsbruxism events to be detected with high reliabilityand reproducibility.

References

1. Rugh JD, Harlan J. Nocturnal bruxism and remporo-mandibular disorders. In; Jankovic J, Tolosa E ¡eds).Advances in Neurology, vol 49. New York: Raven Press,1988:329-341.

2. Ware jC, Rugli JD. Destructivt briixiim: Sleep srage rela-tionship. Sleep 19R8ai:172-liil .

3. Kydd WL, Daly C. Duration of nocturnal tootli contactsduring bruxirg. J Prosrhet Dent 198 5;53:717-721.

4. Reding GR, Zepelin H, Robinson JE Jr, Zimmerman SO,Smith VH. Nocturnal teeth-grinding: All-night psy-chophysiobgic studies. J Dent Res 1968;47:78é-797.

5. Satoh T, I-larada Y. Electrophysiological study on rootli-grinding duting sleep. Electroencephalogr Clin Neuro-physiol 1973;35:267-275.

Robinson JE, Reding GR, Zepelin H, Smith VH, Zim-merman SO. Nocturnal teeth-grinding: A reassessment fordentistry. J Am Dent Assoc 1969;78:1308-1311.Solberg WK, Clark GT, Rugh JD. Nocturnal eleccromyo-graphic evaluation of bruxism patients undergoing shortterm splint therapy. J Oral Rebabil l975;2:215-223.Rugh JD, Solberg WK. EleLtioniyographic studies of brux-ist behavior before and after treatment. J Cal Dent Assoc1975;3:5Ê-S9.

Rugh JD, Johnson RW, Temporal analysis of nocturnalbruxism during EMG feedback. J Periodontol 1981;52:263-265.Clark GT, Beemsterboer PL, Solberg WK, Rugh JD.Nocturnal electrotnyographic evaluation of myofacial paindysfunction in patients undergoing occlusal splint therapy.J Am Dent Assoc 1979;99:607-611.Clark GT, Rush JD, Han de I man SL, Nocturnal massetermuscle aciiviiy and urinary catecholamine levels in brux-ers.J Dent Res 1980;59:1571-1576.Clark GT, Beemstetboet PL, Rugh JD. Tbe treatment ofnocturnal bruxism using contingent EMG feedback v/ithan arousal task. Behav Res Ther 198l;19;451^i5.Clark GT, Beemsterboer PL, Rugb JD. Nocturnal massetetmuscle activity and the symptom of masticatory dysfunc-tion.J Oral Rebabil 198iÍ8:279-286.

Wagner MT. Controlling nocturnal btuxism through tbeuse of aversive conditioning during sleep. Am J Clin Bio-feedback 19Sl;4:S7-92.Piccione A, Coates TJ, George JM, Rosenthal D, Karz-mark P. Nocturnal biofeedback for nocturnal bruxism.Biofeedback Self-Regulation 1S82;7:4O5-419.Stock P, Clarke NG. Monitoring bruxism. Med Biol EngComput 1983;21:295-3OO.Okeson JP, Phillips BA, Berry DTR, Cook Y, Paesam D,Galante J. Nocturnal bruxing events in healthy geriatricsubjects. J Oral Rehabil 1990;17:411-418.Velly Miguel AM, Montplaisir J, Rompre PH. Lund JP,La vigne GJ. Bruxism and othei orofacial movements dur-ing sleep. J Craniomandib Diiord Facial Oral Pain 1992;6:71-81.

Kondo K, Clark GT. A method for detecting a bruxismevent [abstract 1969). J Dent Res 1991;70:512.Okeson JP, Phillips BA, Berry DTR, Baldv/in RM. Noc-turnal bruxing events: A report of normative data and car-diovascular response. J Oral Rehabil 1994;21:623-630.Viimann A, Mailer E, Wildschiodtz G. A system for analy-sis of sleep and nocturnal activity in craniomandibularmuscks.J Orofacial Pain !994;8:2é6-277.Clarke NG, Townsend GC, Carey SE. Bruxing patterns inman during sleep. J Oral Rehabil 19S4;I1:123-127.

Joutnal of Orofacial Pain 281

ikeda et al

Resumen

Criterio Para La Detección Del Bruxismo En HumanosAsociado Con Ei Sueño

Durante ei periodo dei sueño, la electro m ¡og rafia de superficieen el múscuio maselero y la ieciura dei eiectrocsrdiograma dela actividad dei corazón fue ilevada a cabo en nueve individuosque padecen de disfuncion oral motora tbruxismo). En ei propioambiente del liogar de esos individuos, las señales fueron nioni.toreadas durante cuatro noches consecutivas Una informacióntotal de 36 noches fue analizada pare llevar a cabo io siguiente(J) Descnbir ia naturaleza y magnitud de la actividad eiectro-miográfica total dei múscuio rnasetero por encima de ur umbralmínimo de 3% para cada nivel de contracción máximo estable-cido para cada individuo; Í2) Describir los cambios de valor delelectrocardiograma tusando el interuaio r.r) que ocurrieron enrelación a esos cambios electromiograficos. A partir de iosdatos obtenidos se estableció un criterio para ia detección deios eventos bruiistas. combinados en un evento de detecciónaigon'tmico compietamente automático B número promedio y ladetección de ios eventos bruxistas fueron reportados Se pre-senta y se discute la iógica bajo ei criterio seleccionado y elefecto, que otro posible criterio podría tener de la separaciónde los eventos motores normales de lo anormal.

Zusammenfassung

Kriterien für die Entdeckung von Bruxismus vushrend desSchiafes bei Menschen

Oberfiachen-Eiei<tromyographie des Masseters undElektrocardiographieaufnaiirnen wurden wahrend des Schiafesvon neun an Oral-Motorslörungen (Bruxismus) während desSchiafes leidenen Personen aufgenommen. Die Daten wurdenin ihrer normaler Scbiafumgebung zu iHause in vier aufeinander-folgerden Nachten regestriert, insgesamt wurden Daten von 36Nächten anaiysiert, um das Foigende wahrzunebemen: (!) DieBeschreibung der Eigenschaften und des Umfangs dergesamten Wirksami<eit der Masseterelektrographie, die eineWinimaischweiie von 3% der vorher gemessenen maxinialvoiun-laristisclien iHöhe der Muskelveriiürzung bei jedem Individuumüberschntten hat, und (2) Die Beschreibung der Änderungen imGrad der Eiectrocardiograpbie (im Gebrauch des R-R fntervai).die im Vergleich zu diesen electromyographisehen Erhöhungenvorkamen. Kriterien für die Entdeckung der einzelnenBruxismusergebnisse wurden aus diesen Daten formuliert undin einen völlig automatisierten Aigorytbmus kombiniert, der dasErgebnis identifiiieren kann Die durchschnittliche Zahl undDauer des ßruxismusergebnisses werden hier beschrieben Diezugrundeiiegende Logik für die gewähiten Knterien und dieWirkung, die die anderen mogiichen Kriterien der Disknmi-nierung von normaien und unnormalen Motorergebnissen habenwurden, werden ebenfaiis dargesteilt und diskutiert.

282 Volume 10, Number3, Í996