Sample and study area

A representative random sample of 7008 women and men (aged 20 to 79 years) was drawn from three district towns Greifswald, Stralsund, Anklam and 29 communities in the surrounding region, which is part of West Pomerania and is the most northeasterly region of Germany. West Pomerania is the northeastern part of Mecklenburg-West Pomerania, one of the 16 federal German states. The sample selection was done in two steps. First, of the three districts in the region, three cities (17,076 to 65,977 inhabitants) and 12 towns (1516 to 3044 inhabitants) were selected, and of the small towns (less than 1500 inhabitants), 17 out of 97 were drawn at random. Second, from each of the selected communities, subjects were drawn at random from official inhabitant data files proportional to the population size of each community and stratified by age and sex. From the entire population of 212,157 inhabitants, 7008 subjects were sampled, with 292 persons of each sex in each of the 12 five-year age strata. The examinations took place from October 1997 to May 2001. For a full description of the design of the study see John et al3 and for the dental part see Hensel et al.4

Subjective TMD symptoms

To obtain a subjective evaluation of dysfunction symptoms, each participant was interviewed concerning recent unilateral or bilateral symptoms of dysfunction, such as joint sounds like clicking or crepitation, pain in the temporomandibular joints, and pain in the facial muscles (each: sometimes, frequently, always), according to the TMD diagnosis guidelines of the Academy of Orofacial Pain.5 TMD symptoms and their prevalences are shown in Table 1. Together with prevalences of TMD signs, the prevalences of TMD symptoms have been described in detail elsewhere.6 The dependent variable was defined the way that subjects with one or more symptoms belonged to the TMD symptom population. The distribution of TMD symptoms is presented in Table 2.

TABLE 1.  Percentage Distribution of Prevalences of Subjective Symptoms of TMD in 4290 Adults Examined by Means of the Dental Interview

View Fullscreen

TABLE 2.  Distribution of TMD Symptoms^a

View Fullscreen

Normal occlusion/malocclusions

Subjects with a defined loss of teeth (15+ teeth) were excluded from the analysis because it was not possible to adequately assess their morphologic occlusion (for full definition see Gesch et al1). Thirty-three variables of malocclusion and normal occlusion were clinically investigated (Table 3). An anatomically correct occlusion was considered as normal (ideal) occlusion. A deviation from normal occlusion was regarded as malocclusion.7

TABLE 3.  Results of the Univariate Logistic Regressions and Prevalences (%) of Independent Variables. Regressions of the Dependent Variable TMD Symptoms in Dependence on Occlusal Factors (Normal or Malocclusions and Factors of Functional Occlusion) as well as on Parafunctions, Adjusted for Sex, Age, and Socioeconomic Status^a

View Fullscreen

Functional occlusion factors

Fourteen factors of functional occlusion comprising occlusal contacts or interferences during mandibular movement and dental attrition were clinically examined (Table 3). Tooth contacts were analyzed by themselves and in combination with wear facets.8 This was done to see whether the person concerned “uses” the “interfering” occlusion (Table 3). Whether tooth contacts occurred on one or both lateral movements of the mandible as well as whether they occurred on one or both sides on protrusion were registered. Parafunctions like sometimes or frequent grinding or clenching were obtained by the dental part of the online interview (Table 3).

Socioeconomic status

Sociodemographic parameters like age, sex, education, and professional training were determined from the general question section of the online interview. Socioeconomic items such as net income were registered by means of a self-administered questionnaire (Table 3, for definitions see Gesch et al1).

When subjects stated a history of “inflammatory joint diseases” (such as chronic polyarthritis) or “one or more accidents involving cranial injury in the past 12 months” in the medical questions section of the online interview, they were excluded (n = 316).

Clinical examinations were performed by eight calibrated examiners. Prebaseline orthodontic certification of the dentists was based on the examination of 30 pairs of dental casts showing complex symptoms of occlusal deviation that was repeated after several days. The findings recorded by the specialist were the gold standard. Intraexaminer and interexaminer agreement were tested by Cohen's ;gk.9

The results were within the range of “strong agreement” (;gk 0.66–0.81). Regarding examiner agreement for the clinical examinations, the examiners were compared for corresponding prevalences of the recorded malocclusions and the results presented in biannual Data Safety and Monitoring Committee reports. In case of distinct deviation, the examiner concerned was recalibrated. The dental interview was conducted by two trained dental assistants. With the informed consent of the subjects, the data input quality was tested by means of tape recordings and comparison of the acoustic and computerized data in a 10% sample of all interviews. In 18 controlled tape recordings, only one input error per dental assistant was found in 1217 single inputs per subject. For further description of quality management procedures see Hensel et al.410 All measurements and evaluations were performed blinded with respect to the identity of the study participants.

Statistical analysis

The data obtained were recorded online on a standardized data entry sheet and later statistically processed, edited, and analyzed at the Institute of Epidemiology and Social Medicine at the University of Greifswald, using the SAS System (SAS Institute Inc., Cary, NC). Statistical test results with P < .05 are considered statistically significant.

At stage 1, potential occlusal factors for TMD symptoms were screened with several multivariate logistic regression models (SAS PROC LOGISTIC) including age, sex, socioeconomic factors, and one of the occlusal factors in each model (thus further called “univariate models”). Also interactions of occlusal factors with sex were tested to detect possible risk differences between sexes. Results and prevalences of occlusal factors and other factors are shown in Table 3. At stage 2, all independent variables and interactions between sex and occlusal factors were simultaneously entered into a multivariate logistic regression model. With the stepwise backward method, first all nonsignificant interactions and afterward all nonsignificant single independent variables were eliminated (SAS SLSTAY = 0.05, SLENTRY = 0.20). The results were adjusted for sociodemographic and socioeconomic variables as in the univariate models. Intervariable correlations were examined for potential multicollinearity between variables.

In this area of research it has been recommended that, for an odds ratio (OR) to be relevant, it would need to be statistically significant and, for it to be clinically noticeable, it should represent a doubling (OR > 2)1112 or halving (OR < 0.5) of risk, although prevalence of disease and base risk must be taken into account for a final decision of clinical relevance. An OR of less than 1 indicates that the presence of the factor is associated with reduced risk, an OR more 1 indicates increased risk.

Random sample

Of the 7008 subjects initially sampled, 741 dropped out, either because they moved away (n = 615) or because of death (n = 126), reducing the sample size to 6267. A response rate of 68.2% for men and 69.4% for women, resulted in a sample of 4310 subjects. The overall response rate was 68.8% for the ages of 20 to 81 years and 71.3% for the age groups 20 to 74 years. An analysis of nonresponders found that the main reasons for nonparticipation were disinterest (39.7%), health problems (23%), adequate available medical care (11.6%), lack of time (16.7%), fear of examination results (3%), and others (6%).4 Twenty subjects were unable to undergo the oral examination. Finally, 4290 adults could get investigated anamnestically for symptoms of temporomandibular dysfunction and clinically for different types of malocclusions as well as for factors of functional occlusion. Sex and age distribution of the sample are given in Table 4.

TABLE 4.  Sex and Age Distribution of the Sample (n = 4290)

View Fullscreen

Results of the univariate logistic regression analysis

The univariate analysis revealed that the dependent variable TMD symptoms were associated with one factor of morphologic occlusion (spacing) and parafunctions (clenching, grinding) but none of the factors of functional occlusion. It was not possible to calculate any ORs for unilateral or bilateral open bite above three mm because of the small number of observations (Table 3).

Interactions between sex and a single occlusal factor were not significant throughout, apart from unilateral non– working side interference and unilateral scissors-bite (total buccal crossbite), but in these cases the occlusal factor was not significant itself. An exception was the variable spacing (meaning primary spacing, not because of tooth loss).

Results of the multivariate regression analysis

Although grinding was significantly associated with TMD symptoms in the univariate logistic regressions (P < .001), it proved nonsignificant when considered simultaneously with the other variables of the multivariate logistic regression (Table 5). On the other hand, wear facets in dental restorations were not significant in the univariate, but in the multivariate logistic regression (OR = 0.7, Table 5; for a detailed description of the prevalences of wear facets see Bernhardt et al13). As in the univariate logistic regressions, the dependent variable TMD symptoms were significantly associated with spacing with an OR of 0.6 (prevalence 27.3%), with frequent clenching (OR = 3.4, 5.2%), and with the sex by spacing interaction. Only subjects who were aware of frequent clenching, more frequently stated one or more subjective TMD symptoms compared with those without this parafunction.

TABLE 5.  Results of the Multivariate Logistic Regression. OR and 95% CI for the Dependent Variable TMD Symptoms in Dependence on Occlusal Factors (Normal or Malocclusions and Factors of Functional Occlusion) as well as on Parafunctions, Adjusted for Sex, Age, and Socioeconomic Status^a

View Fullscreen

On the other hand, the occlusal factors spacing and wear facets in dental restorations were connected with less frequently perceived TMD symptoms. Normal occlusion was not significantly associated with TMD symptoms. The same was true for the sociodemographic and socioeconomic parameters. The ORs of none of the significant variables of the multivariate logistic regression analysis were between one and two. They were either below or, solely in the case of the nonocclusal variable frequent clenching, higher than two.

The only significant sex by spacing interaction showed a differential association between spacing and TMD symptoms for men and women. Women with this morphologic occlusion showed less frequent symptoms of TMD compared with those without spacing. All intervariable correlations between the occlusal variables of the final multivariate logistic models were <0.25 (Pearson), showing no potential redundancies.