76 THE JOURNAL OF PROSTHETIC DENTISTRY VOLUME 83 NUMBER 1

Despite decades of clinical experience and debate,there is persistent confusion about the relationship ofocclusion to the identification and treatment of tem-poromandibular disorders (TMDs). A lot of the con-troversy on this subject has resulted from simplistic uni-variate examination of occlusal variables,1 whereas inbiologic models, single factors rarely act in isolation.The influence of any single structural occlusal variableis always contingent on what else is coacting. Thus,multifactorial diseases such as TMD require a multifac-torial approach in analysis in which multiple factorsmust be considered simultaneously. However, there are

few of these studies. From the few traditional stepwisemultiple logistic regression models,2,3 the amount oflog likelihood accounted for by occlusal variables isestimated to range at most from approximately 5% to27%,1 which supports a limited contribution of occlu-sion to the differentiation of TMD patients. However,this also implies that 73% to 95% of TMD patient char-acteristics were not accounted for by the occlusal fea-tures studied.

The occlusal debate was recently highlighted in aneditorial4 and in a position paper5 through which theAmerican Equilibration Society (AES) responded to aperceived diminution of importance of occlusion in thediagnosis of TMD.6-9 These articles sponsored by theAES4,5 correctly point out several items of consensualagreement, for example that diagnostically the “use ofAngle’s classification to define ‘malocclusion’ is perhapsthe most serious…flaw in literature…,” and that “both

Analysis of occlusal variables, dental attrition, and age for distinguishinghealthy controls from female patients with intracapsular temporomandibulardisorders

Donald A. Seligman, DDS,a and Andrew G. Pullinger, DDS, MScb

UCLA School of Dentistry, Los Angeles, Calif.

Statement of problem. Confusion about the relationship of occlusion to temporomandibular disorders(TMD) persists.Purpose. This study attempted to identify occlusal and attrition factors plus age that would characterizeasymptomatic normal female subjects.Methods and material. A total of 124 female patients with intracapsular TMD were compared with 47asymptomatic female controls for associations to 9 occlusal factors, 3 attrition severity measures, and ageusing classification tree, multiple stepwise logistic regression, and univariate analyses. Models were tested foraccuracy (sensitivity and specificity) and total contribution to the variance. Results. The classification tree model had 4 terminal nodes that used only anterior attrition and age.“Normals” were mainly characterized by low attrition levels, whereas patients had higher attrition and tend-ed to be younger. The tree model was only moderately useful (sensitivity 63%, specificity 94%) in predictingnormals. The logistic regression model incorporated unilateral posterior crossbite and mediotrusive attritionseverity in addition to the 2 factors in the tree, but was slightly less accurate than the tree (sensitivity 51%,specificity 90%). When only occlusal factors were considered in the analysis, normals were additionally char-acterized by a lack of anterior open bite, smaller overjet, and smaller RCP-ICP slides. The log likelihoodaccounted for was similar for both the tree (pseudo R2 = 29.38%; mean deviance = 0.95) and the multiplelogistic regression (Cox Snell R2 = 30.3%, mean deviance = 0.84) models.Conclusion. The occlusal and attrition factors studied were only moderately useful in differentiating nor-mals from TMD patients. (J Prosthet Dent 2000;83:76-82.)

aAdjunct Associate Professor, Division of Oral Biology and Medi-cine, Section of Orofacial Pain.

bProfessor, Division of Oral Biology and Medicine, Section of Oro-facial Pain; Director Residency Program in Orofacial Pain andDysfunction.

CLINICAL IMPLICATIONS

The results of this study indicated a patent but limited association of occlusal compo-nents to TMD, and a lack of extreme occlusal variation in most normals. Some mea-surement extremes can obviously be tolerated as normal depending on the other struc-tural contingencies. Because occlusal effects are contingent on highly variable andindividual interactions, clinicians should exercise caution before indicting or rejectinga particular occlusal feature as important in a TMD patient’s disease.

careful clinical observation and logic would showthat…progressive anterior open bite and overjet…andprogressive asymmetry…[are] automatic conse-quence(s) of lost condylar height….” However, severalother belief systems concerning the relationship ofocclusion to TMD were also included that still requirevalidation through scientific investigation.

Controlled studies to properly test the debatedocclusal issues are welcome because the decades ofanecdotal clinical experience cannot be rejected out ofhand. Meanwhile, in some circles, there has been areturn toward the advocacy of occlusal adjustmenttreatment in children as a prophylactic measure,10-13

despite a failure to control behavioral influences in thereported mild reduction of muscle hyperirritability inchildren with only TMD-like symptoms rather thanTMD. Proof of the efficacy of occlusal therapy wouldserve many TMD patients well if the profiles appropri-ate to occlusal or structural change were identified,while avoiding overtreatment in profiles that wereunrelated.

Unfortunately, the efficacy of and guidelines forpatient selection have not been published. This isimportant because the prophylactic experimentalmodel in children shows roughly the same limitedstrength of relationship to occlusal factors as in themultiple regression prevalence-based models in adults,1which implies that many potentially more importantcofactors remain to be addressed. Meanwhile, the 1983ADA Guidelines state that irreversible therapies shouldbe restricted to patients whose acute symptoms havebeen resolved and that occlusal adjustment therapyshould not be used routinely to treat TMD patients orprophylactically.14

Earlier univariate analyses are limited in their abilityto identify health or disease.1 Their failure is mainlyconcerned with sensitivity,1 signifying a problem differ-entiating normals from TMD patients. The nullhypotheses of this study was an extension of one testedin earlier studies,1,2 namely, that occlusal relationshipsplus age and dental attrition are not able to differenti-ate female patients with intracapsular TMD from nor-mal subjects. The primary purpose of this study was theidentification of occlusal and attrition factors plus agethat would characterize asymptomatic normal subjectsfrom grouped TMD patients with intracapsular disor-ders. It is not our purpose to use occlusion as a test forTMD; that would be better accomplished by clinicalexamination. Nonetheless, identification of normativemodels of occlusion that are associated with less dys-function could be useful in many aspects of dentaltreatment.

MATERIAL AND METHODS

Subjects with complete occlusal data were selectedfrom a set of patients and controls used in a previous

study on dental attrition.15 Subjects were limited to asubset of female patients and controls because femalesare the majority of persons seeking treatment. Theaddition of dental attrition and age broadened thescope of multifactorial analysis that was previously1,2

limited to occlusal structural variables.For the experimental sample, 124 female patients

with intracapsular TMD (mean age 35.4 ± 11.89 years;range 13-72 years) with a complete occlusal and attri-tion factor data set were selected retrospectively fromthe pool of 239 previously differentiated clinicalcases.15 For the control sample, 47 of the 51 femaleasymptomatic control subjects (mean age 41.2 ± 15.48years; range 21-74 years) used in a previous study15

were selected because they had a complete occlusal fac-tor database plus attrition measurements.

The inclusion criteria included a thorough clinicalexamination to confirm the presence of TMD accord-ing to the guidelines of the American Academy of Oro-facial Pain (AAOP).16 The subjects were assigned to 2diagnostic groups that conformed to the classificationcriteria16: (1) TMJ disk displacement (both reducingand nonreducing: n = 51), and (2) TMJ osteoarthrosis(n = 73). The diagnostic groups were then pooled forcomparison to the asymptomatic controls.

For measurement, previously published data collec-tion techniques were used.2,15 The 12 occlusal andattrition features considered were as follows: (1)retruded-contact position (RCP) to intercuspal-posi-tion (ICP) slide length; (2) overbite; (3) overjet; (4)unilateral posterior crossbite; (5) anterior open bite; (6)incisor dental midline discrepancy; (7) number of unre-placed missing posterior teeth; (8) first molar mesiodis-tal relationship; (9) right and left first molar positionasymmetry; (10) anterior dental attrition severity; (11)mediotrusive dental attrition severity; and (12) lat-erotrusive dental attrition severity. With the inclusionof age, there was a total of 13 independent variablesincluded in the analysis. Occlusal factors included allfactors used in the previous study,2 with the exceptionof unilateral contact in RCP and asymmetrical RCP-ICP slide, which were not recorded in the new sampleand which were not significant in the earlier study.

To ensure blinding and bias controls, every effortwas made so both the patient assignment and the datacollection were performed in an unbiased manner. Forocclusal measurements from study casts, the identity ofthe subject was unknown and the data measurementwas collected by a single examiner.

Statistical analysis

Univariate comparisons were made for the occlusalvariables using the Fisher 2-tailed exact test and chi-square analysis with a Yate’s correction for nominalvariables, the Mann-Whitney U-test for ordinal vari-ables, and 1-way analysis of variance (ANOVA) for the

SELIGMAN AND PULLINGER THE JOURNAL OF PROSTHETIC DENTISTRY

JANUARY 2000 77

interval variables. The level of significance was set atP≤.05.

The goal was the identification of factors associatedwith health. The classification tree method (SPSSAnswer Tree 1.0 software, SPSS Inc, Chicago, Ill.)17

was used to search for hidden structure in the data, toidentify interactions, and to reveal bipolar relationships.The model is fitted using binary recursive partitioning,whereby the data are successively split along coordinateaxes of the predictor variables so that at any node, thesplit that maximally distinguished the response variablein the left and the right branches is selected. Splittingcontinues until nodes are pure (all observations are thesame class in a node) or data are too sparse (default: atleast 5 in the daughter node).18 The predictionassigned to each terminal node is for that class (asymp-tomatic vs patient) with a significant increase in repre-sentation when compared with its overall prevalence inthe entire sample. For comparison purposes, a classifi-cation tree analysis was also performed for the 9occlusal variables without age and attrition variables.The 2 outcome variables in the trees were always dis-ease or health.

The impurity measure used for categorical targetswas “Gini” and the minimum change in impurity wasset at 0.0001. Prior probabilities were assumed to beequal for both classes. To avoid overfitting, trees werepruned based on the Standard Error Rule (1.0 multi-plier). Overestimates of R2 and risk estimates were con-trolled by partitioning the data into a training samplefor construction of the tree and a testing sample. Riskestimate and percentage reduction in the maximum loglikelihood (“Pseudo-R2”) and its accuracy (meandeviance; accurate ≤1.0) were determined from the

testing sample outcome. The ability of the classificationtrees to predict health was tested by tabulation of thesensitivity and specificity of the testing samples.

To arrive at the tree that best represented the sam-ple, 10 trees were generated, each using a different70%/30% partitioning of the entire sample. In all 10instances, the only variables that were used were ante-rior attrition and age. Fifteen additional partitionedtrees were generated using only these 2 variables toidentify the tree pattern that was most stable. One pat-tern appeared in 10 of the 15 trials, and this was select-ed as the most representative tree. The accuracy of thetree was computed as the percentage classified incor-rectly for the 30% partitioned setaside test samples ofthe 10 trees with the representative pattern. The sensi-tivity and specificity were similarly computed. Thepseudo-R2 and mean deviance were computed fromthe test sample distribution of the composite represen-tative tree.

The classification tree analysis was used to identifypotential interactions for inclusion in a complementarymultiple stepwise logistic regression analysis (SPSS Base8.0 software, SPSS Inc). All variables and interactionssuggested by the tree were entered into the logisticmodel. Selection was made among the potential predic-tors using backward stepwise selection methods. A P≥.10was required for removal; P≤.50 was required for reentry.

With the results from the multiple logistic regressionanalysis, the Odds Ratio for disease was simultaneouslyassessed for each significant occlusal factor. The OddsRatio is a measure of association without etiologicimplication that describes the proportionate risk that anindividual in the sample with a certain occlusal featurewill belong to a particular group, while simultaneously

THE JOURNAL OF PROSTHETIC DENTISTRY SELIGMAN AND PULLINGER

78 VOLUME 83 NUMBER 1

Table I. Univariate comparisons of occlusal variables

Patients Controls

Mean† Range Mean† Range Differences

RCP-ICP slide 0.77 0-5 0.41 0-2 P<.008*Overbite 2.81 –3.5-15 3.12 0-10.0 NS*Overjet 2.73 0-10 2.46 0-5 NS*Anterior open bite 10/124 0/47 NS**Midline discrepancey 0.94 0-4.5 0.68 0-3 NS*Missing teeth 0.93 0-5 0.64 0-4 NS*First molar relation –0.69 –7.5-+8 –0.81 –6-+5 NS*R/L asymmetry 1.09 0-8 1.05 0-6 NS*Crossbite 20/124 3/47 P<.01***Anterior attrition 7.55 0-12 5.28 2-8 P<.001****Mediotrusive attrition 8.06 2-16 8.13 2-14 NS****Laterotrusive attrition 17.37 2-32 16.98 6-30 NS****

NS - Not significant.*1-way analysis of variance.**Fisher’s 2-tailed exact test.***Chi-square test with Yate’s correction.****Mann-Whitney U-test.†Frequencies noted for anterior open bite and unilateral posterior crossbite.

controlling for all the other occlusal variables. A 1:1ratio implies no relative risk for disease. The ratios forbinary predictors are single odds ratios. Odds ratios forinterval variables are for a 1-unit change in the variable,and increase geometrically for each unit increase in themeasurement. An estimation of the total log likelihood(Cox Snell R2)19 and its mean deviance explained by asummation of the significant occlusal factors wasobtained.

SELIGMAN AND PULLINGER THE JOURNAL OF PROSTHETIC DENTISTRY

JANUARY 2000 79

RESULTSUnivariate comparisons of occlusal factors

Anterior open bite was demonstrated in 8% of thepatients and 0% of the controls (Table I). Unilateralposterior crossbite was more common in the patients(P<.01), and length of the RCP-ICP slide was larger inthe patients (P<.008). Anterior attrition was moresevere in the patients (P<.001).

Fig. 1. Classification tree derived from all 13 occlusal, attrition, and age variables for pre-dicting normal controls (Asx) and TMD patients (Pt). Tree begins with anterior attrition sever-ity and follows arrows downward until terminal node is reached. For example, endpoint forpathway 2 is reached with anterior attrition severity score smaller than 8.0 and age above37.5 years. Sample sizes and class are shown for each terminal node for testing sample alone.

Fig. 2. Classification tree derived from only 9 occlusal variables for predicting asymptomaticnormal controls (Asx) and TMD patients (Pt). Tree begins with anterior open bite and followsarrows downward until terminal node is reached. For example, endpoint for pathway 3 pre-dicting asymptomatic normal control is reached with absence of anterior open bite, overjetsmaller than 5.25 mm, and RCP-ICP slide length smaller than 1.75 mm. Sample sizes andclass are shown for each terminal node for testing sample alone.

Classification tree analyses

With all 13 factors considered, a classification treewas formulated with 4 terminal nodes (Fig. 1). Path-ways 1 and 4 identified predominantly disease, whereaspathways 2 and 3 predominantly identified asympto-matic controls (healthy normals). Only anterior attri-tion and age were used in the tree, whereas the 9occlusal factors and both laterotrusive and mediotru-sive attrition were not useful. The classification treeaccounted for 29.4% (pseudo R2) of the log likelihood(mean deviance = 0.953). The tree was accurate inassigning 69% of the testing sample (31% risk estimate,SE = 0.068) to the correct group. The interaction of“anterior attrition severity × age” was consequentlyadded to the list of factors used later in the multipleregression analysis.

When only 9 occlusal factors were considered, a clas-sification tree was formulated with 4 terminal nodes(Fig. 2). Pathways 1, 2, and 4 identified predominant-ly disease; pathway 3 predominantly identified asymp-tomatic normals. Only anterior open bite, overjet, andthe RCP-ICP slide length were used in the tree. Theclassification tree was able to explain 11.4% (pseudoR2) of the log likelihood (mean deviance = 1.165). Thetree was accurate in assigning 62.5% of the testing sam-ple (37.5% risk estimate, SE = 0.036).

Multiple stepwise regression analysis

In comparison, the multiple stepwise logistic regres-sion model of the full data set incorporated anteriorattrition severity, age, mediotrusive attrition severity,and crossbite (Cox Snell R2 = 30.3%, mean deviance =0.84) (Table II). When limited to 9 occlusal factorsomitting age and attrition, only RCP-ICP slide lengthremained in the model (Cox Snell R2 = 4.8%, meandeviance = 1.14). The full data set regression modelwas able to correctly assign 78.5% of the total sample(sensitivity = 51%; specificity = 90%).

Identification of normal subjects

The full data set classification tree model (Fig. 1) wasonly marginally accurate in identifying healthy persons

(n = 47): sensitivity 63%, specificity 94%. Forty-six percentof the controls were characterized by an anterior attritionlower than the patient average (pathway 3). The remain-ing controls were older patients having average or lessanterior attrition (pathway 2), and 37% of the patientswere falsely identified as normals (pathways 2 and 3).

The occlusal factor classification tree (Fig. 2) wasaccurate in identifying healthy individuals (specificity =100%), all of whom were characterized by an absence ofanterior open bite, smaller overjet (<5.25 mm), andsmaller RCP-ICP slide lengths (<1.75 mm).However,72% of patients were misclassified as normal by theseparameters (pathway 3; sensitivity 25%).

The multiple logistic regression model also associat-ed the normal sample with lower anterior attritionseverity scores and an older mean age than the patients(Table I). In addition, the normals were characterizedby the absence of unilateral posterior crossbite.Although weakly significant, the mediotrusive attritionscore did not approach a 2:1 odds ratio for clinical sig-nificance within the scoring range (Table I). A 2:1 oddsratio would have required a score of 32, but the maxi-mum scoring range was only 16. Thus, this cofactor hasdoubtful clinical relevance.

DISCUSSION

This study shows that it was only moderately possi-ble to differentiate a female sample of asymptomaticnormals from patients attending for treatment ofTMDs with a classification tree (Fig. 1) or multipleregression (Table I) analyses. The deficiency is eitherlow sensitivity (61.3%, classification tree analysis) orlow specificity (51.1%: multiple logistic regressionanalysis).20 This implies a major overlap of occlusal andattrition parameters among patients and normals aspreviously reported.2,15 Subjects with benign TMDsymptoms not seeking evaluation or treatment werenot discussed because they were not part of the sample.However, we would assume that persons with mildersymptoms would be even more difficult to typify.

Occlusal characteristics of the normals (controls)were mainly characterized by anterior attrition below

THE JOURNAL OF PROSTHETIC DENTISTRY SELIGMAN AND PULLINGER

80 VOLUME 83 NUMBER 1

Table II. Significant associates from multiple logistic regression: Odds ratios and values to reach a 2:1 odds ratio for diseaseidentification

Factor β P= Odds ratio at the mean Value to reach 2:1 odds ratio

Full data setAge –0.453 .0090 1.22:1 <33 yCrossbite 1.7636 .0239 11.67:1 PresenceAnterior attrition 0.8288 .0001 6.57:1 7.7 scoreMediotrusive attrition –0.3271 .0047 1.02:1 32 score

Occlusal factors onlyRCP-ICP slide length 0.7944 .0100 1.33:1 3.15 mm

the patient average. However, average or less anteriorattrition is also common in patients, making distinctionproblematic at the lower attrition ranges (Fig. 1).Although attrition would logically be expected to wors-en with age, the relationship between the two is com-plex and nonlinear.21 In our study, the multiple regres-sion analysis in its linear construct automatically con-trolled for the influence of age on the attrition level,and identification of normals became more probable asthe age increased with a decreased anterior attrition andabsence of crossbite (Table II).

The combined results from the 3 types of analyses(classification tree, logistic regression, and univariate)suggest that asymptomatic controls were characterizedby low levels of anterior attrition severity, smaller or noRCP-ICP slides (<1.75 mm), absence of extreme over-jet (>5.25 mm), and absence of unilateral posteriorcrossbite. However, because either the sensitivity or thespecificity did not reach target ranges (sensitivity >75%;specificity >90%),20 reliance on these features for diag-nosis is problematic. When one allows occlusal variablesto compete in a predictive model with attrition, whichis an acquired change or an adaptation of structuresthat may be behavioral in origin, most occlusal factorsare excluded by attrition. The strength of attrition incomparison to occlusal factors is a new and interestingfinding, especially if it could be corroborated in anoth-er completely separate sample of subjects and controls.

It may be useful to typify a normative occlusion asthat associated with the lowest risk for codevelopmentof TMD problems, but it is probably inappropriate toapply those parameters to reverse an intracapsular prob-lem once established.22 For example, correction of uni-lateral functional crossbite might reduce the adaptationdemands in a growing adolescent, but may be inconse-quential in a symptomatic adult because skeletal adap-tation has occurred and correction is unlikely to reversean intracapsular problem.21 Similarly, although correc-tion of some occlusal anomalies through periodicocclusal equilibration in children and teenagersappeared to reduce the development of mild TMD-likesymptoms in developing persons,10-13 the same benefitwas not seen in equilibration of young adults.23

Because the few occlusal factors that characterizedpatients (Fig. 2) disappeared when nonocclusal factorswere added to the model (Fig. 1), a presumption ofocclusal cause and a dental occlusal treatment approachmight be ignoring more important contributing factorsin many patients. Contemporary clinical guidelines7,8,14

suggest a needs assessment for occlusal treatment shouldbe deferred until symptoms are controlled.

Regarding limitations of this study, the logisticregression analysis and classification tree analysis com-plement each other, and allow clinicians to betterunderstand the factors on which to concentrate theirattention. It must be conceded that, whereas we have

SELIGMAN AND PULLINGER THE JOURNAL OF PROSTHETIC DENTISTRY

JANUARY 2000 81

validated the results of the previous multiple regres-sion analyses through similar outcomes in severalindependent populations,1,2 the results of the treeanalysis are only validated in this study in a set-asidetest sample and must be validated in a new sample ifits results are to be applied. When studying theocclusal factors without attrition and age, theexplained variance in the multiple regression examina-tion was only 4.8%, which was less than the 4.8% to27.1% reported in the earlier regression study.1 Theexplained log likelihood increased to 30.3% (Cox SnellR2) with the addition of age and attrition severity,which was similar to the 29.4% for the correspondingmeasurement of explained variance (pseudo R2) forthe classification tree. Although this level of contribu-tion is not zero, at least two thirds of the contributionis unexplained by the considered factors. All iterativeselection procedures, including stepwise selection pro-cedures, have a tendency to overfit the data. This istrue for all observational, exploratory studies. Howev-er, we believe that missing an important predictor atthis stage is worse than falsely including a factor thatwill later prove to be irrelevant.

It must be emphasized that this study was also lim-ited by and conditional to the factors included. Anyindividual factor contribution is likely to be diminishedby the introduction of other factors into the model.Thus, the contribution of factors tested is relative andthe actual contribution likely less than the 29% to 30%reported in our study. The addition of any strongerassociates to TMD would probably even eliminatesome of the factors found significant in this analysis.Thus, it is important to avoid single factor explanationsof these complicated multifactorial problems. Thisspeaks to the importance of a broad-based evaluation ofpatients with TMD problems.

CONCLUSIONS

Within the limitations of this study, the followingconclusions were drawn:

1. The occlusal and attrition features plus age wereonly moderately useful for differentiating the normalsfrom the TMD patients. Therefore, this study affirmeda previously described2 limited but measurable associa-tion of occlusal components to TMD, albeit withoutetiologic implications.

2. Apart from the domain of a few extreme ranges,occlusal models for TMD patients remain elusive.

We thank Dr Jeffrey Gornbein, Department of Biomathematics,School of Medicine, University of California at Los Angeles, for hishelp in suggesting the statistical model and the analysis.

REFERENCES

1. Pullinger AG, Seligman DA. Quantification and validation of predictivevalues of occlusal variables in TMD using a multifactorial analysis. J Pros-thet Dent 2000 [in press].

THE JOURNAL OF PROSTHETIC DENTISTRY SELIGMAN AND PULLINGER

82 VOLUME 83 NUMBER 1

2. Pullinger AG, Seligman DA, Gornbein JA. A multiple logistic regressionanalysis of the risk and relative odds of temporomandibular disorders asa function of common occlusal features. J Dent Res 1993;72:968-79.

3. De Laat A, van Steenberghe D, Lesaffre E. Occlusal relationships and TMJdysfunction. Part II. Correlations between occlusal and articular parame-ters and symptoms of TMJ dysfunction by means of stepwise logisticregression. J Prosthet Dent 1986;55:116-21.

4. Dawson PE. Why NIH is wrong about “TMD.” Cranio 1997;15:1-3.5. Dawson PE. Diagnosis, management and treatment of temporomandibu-

lar disorders (TMD). (Position Paper). Submitted by the American Equili-bration Society to NIDR, NIH 1996:1-14.

6. National Institutes of Health. Technology assessment conference state-ment: management of temporomandibular disorders: April 29-May 1,1996.

7. American Academy of Orofacial Pain, Jeffrey P. Okeson, editor. Orofacialpain: guidelines for assessment, diagnosis and management. 3rd ed.Chicago: Quintessence; 1996. p. 122-3.

8. American Academy of Orofacial Pain. Okeson JP, editor. Orofacial Pain:Guidelines for assessment, diagnosis and management. 3rd ed. Chicago:Quintessence; 1996. p. 153-5.

9. McNamara JA Jr, Seligman DA, Okeson JP. Occlusion, orthodontic treat-ment, and temporomandibular disorders: a review. J Orofac Pain1995;9:73-90.

10. Kirveskari P, Alanan P, Jämsä T. Association between craniomandibulardisorders and occlusal interferences. J Prosthet Dent 1989;62:66-9.

11. Kirveskari P, Alanen P, Jämsä T. Association between craniomandibulardisorders and occlusal interferences in children. J Prosthet Dent 1992;67:692-6.

12. Kirveskari P, Alanen P. Odds ratio in the estimation of the significance ofocclusal factors in craniomandibular disorders. J Oral Rehabil 1995;22:581-4.

13. Karjalainen M, Le Bell Y, Jämsä T, Karjalainen S. Prevention of temporo-mandibular disorder-related signs and symptoms in orthodontically treat-ed adolescents. A 3-year follow-up of a prospective randomized trial.Acta Odontol Scand 1997;55:319-24.

14. Griffiths RH. The President’s conference on the examination, diagnosis,and management of temporomandibular disorders. Chicago: AmericanDental Association; 1983. p. 183.

15. Pullinger AG, Seligman DA. The degree to which attrition characterizesdifferentiated patient groups of temporomandibular disorders. J OrofacialPain 1993;7:196-208.

16. American Academy of Orofacial Pain. Okeson JP, editor. Orofacial pain:guidelines for assessment, diagnosis and management. 3rd ed. Chicago:Quintessence; 1996. p. 19-36, 128-37.

17. Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and regres-sion trees. Belmont: Wadsworth International Group; 1984.

18. Kass GV. An exploratory technique for investigating large quantities of cat-egorical data. Appl Statist 1980;29:119-27.

19. Cox DR, Snell EJ. Analysis of binary data. 2nd ed. London: Chapman andHall; 1989. p. 209.

20. Widmer CG, Lund JP, Feine JS. Evaluation of diagnostic tests for TMD. JCalif Dent Assoc 1990;18:53-60.

21. Seligman DA, Pullinger AG. The degree to which dental attrition in mod-ern society is a function of age and of canine contact. J Orofacial Pain1995;9:266-75.

22. Thurston MH, Turley PK, Pullinger AG. Craniofacial characteristics asso-ciated with a unilateral posterior crossbite in the permanent dentition. JDent Res 1987;66:348 (abstract).

23. Kirveskari P, Le Bell Y, Salonen M, Forssell H, Grans L. Effect of elimina-tion of occlusal interferences on signs and symptoms of craniomandibu-lar disorders in young adults. J Oral Rehabil 1989;16:21-6.

Reprint requests to: DR DONALD SELIGMAN

DIVISION OF ORAL BIOLOGY AND MEDICINE-SECTION OF OROFACIAL PAIN

UCLA SCHOOL OF DENTISTRY

CHS 43-009LOS ANGELES, CA 90024-1668FAX: (323) 665-7048E-MAIL: DonaldS261@aol.com

Copyright © 2000 by The Editorial Council of The Journal of ProstheticDentistry.

0022-3913/2000/$12.00 + 0. 10/1/103746