No plausible explanation for the higher prevalence of TMD in women has been proposed, but it seems reasonable to ask whether the sex hormones may play some role in etiology or maintenance of the condition. For instance, variations in sex hormone levels associated with the menstrual cycle may modulate pain perception.^4–7 Also, prescription of exogenous estrogen in the form of estrogen replacement in menopause or oral contraceptives during reproductive years may increase the risk of being referred for TMD care.⁸ Exogenous estrogen in the form of oral contraceptives does not affect mean pain intensity ratings in regularly ovulating patients with TMD, but it might affect the pattern or variability of pain ratings.⁹ In addition, use of oral contraceptives in healthy ovulating women does not affect the development of TMD-like symptoms elicited by voluntary low-level jaw clenching.¹⁰ These previous studies provide highly mixed support for a possible etiologic or maintenance role for estrogen in the pathogenesis of TMD.

We undertook a study to further elucidate the relationship between exogenous estrogen use and TMD. We confirmed current estrogen use by patients’ self-report and inspection of patients’ medication containers. To minimize self-selection bias owing to voluntary health care–seeking behavior and to maximize the generalizability of results, we used a randomly selected community sample. Signs and symptoms of TMD were quantified through direct physical examination by a calibrated clinical examiner (S.S.) using a standardized assessment measure, the Craniomandibular Index, or CMI.^11,12 The analysis included overall scores and subscores designed to assess muscle and joint signs and symptoms.

	SUBJECTS, MATERIALS AND METHODS

TOP ABSTRACT SUBJECTS, MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
Subjects. Subjects were 510 women 37 to 82 years of age at the time of the study who were participants in the Oral Health: San Antonio Longitudinal Study of Aging, or OH:SALSA, an epidemiologic survey of oral health in a population-based probability sample of Mexican Americans and European Americans living in San Antonio. Subjects represented a stratified random sample of three neighborhoods: a low-income, nearly exclusively Mexican American "barrio" neighborhood; a middle-income, "transitional" mixed Mexican American/European American neighborhood; and an upper-income "suburban" neighborhood containing approximately 10 percent Mexican Americans and 90 percent European Americans. Only Mexican Americans were sampled in the barrio neighborhood because of the small number of European Americans living there. Stratified random sampling was conducted in the transitional and suburban neighborhoods to obtain approximately equal numbers of Mexican American and European American subjects. Pregnant subjects were excluded, as were subjects whose ethnic group affiliation could not be identified as Mexican American or European American according to a standardized algorithm.¹³

Procedure. A research associate completed a home-based assessment and a clinic-based oral examination with each subject. During the home-based assessment, subjects provided information on their health status, patterns of health care utilization, lifestyles and current oral health and function. The home-based assessment also included collection of detailed information concerning the use of prescription drugs, over-the-counter drugs and folk or home remedies. Subjects were asked to show the research associate all the medicines they had taken or used in the preceding two weeks, as well as the drugs that they took or used only as needed. During the clinic-based oral examination, subjects were asked to show the research associate the containers for all medicines, both prescription and over-the-counter, that they had taken in the preceding 24 hours. Subjects were identified as using estrogen if, at either assessment, they stated that they were taking any medication containing estrogen or produced the container for any medication containing estrogen. They were identified as not using estrogen if they failed to produce evidence of estrogen use at both assessments. We compiled a list of all medications currently taken by subjects, and medications containing estrogen were identified by consulting the Physicians’ Desk Reference and a board-certified gynecologist. This method identified 174 women as estrogen users and 336 women as non–estrogen users. We did not record information on dosages of estrogen. Using the same methods, we identified subjects using nonsteroidal antiestrogen, tamoxifen citrate or any medication containing progestin.

A calibrated clinical examiner (S.S.) assessed signs and symptoms of TMD using the CMI.^11,12 The CMI yields an overall score, the CMI score; a Muscle Index, or MI, score, which quantifies pain in facial, neck and shoulder muscles elicited by standard palpation; and a Dysfunction Index, or DI, score, which quantifies temporomandibular joint, or TMJ, dysfunction (pain, noises and mandibular range of motion). The CMI shares considerable methodological and item content overlap with the newer Research Diagnostic Criteria for TMD,¹⁴ which were unavailable when this study began.

Although the study was not specifically designed to blind the CMI examiner to the subject’s estrogen status, the clinical examiner did not have access to information gathered by other examiners during the home-based assessment and the clinic-based oral examinations.

Statistical analysis. Subjects were not systematically asked whether they had reached menopause or had undergone surgical oophorectomy. Therefore, nonovulating subjects could be identified only by their age. The primary analysis included all available female subjects (age range, 37 to 82 years). In addition, to guard against the possibility that the results could be confounded by circulating endogenous estrogens in premenopausal subjects, we conducted secondary sensitivity analyses to determine whether altering the likelihood that subjects were postmenopausal would materially affect the results. We assessed statistical association involving categorical variables using Pearson’s ² and the Fisher exact test. Association of variables recorded on an ordinal or higher scale was assessed using the Spearman rank correlation. We made comparisons of independent groups using the Mann-Whitney test, Savage test and Student t test for independent groups. Linear multiple regression analysis and multiple logistic regression analysis were used to assess the impact of potential confounding variables. Because CMI scores were positively skewed in this nonclinical sample, we transformed CMI scores by taking the square root before statistical testing. All statistical tests were nondirectional, and we rejected all null hypotheses at the .05 level of significance.

	RESULTS

TOP ABSTRACT SUBJECTS, MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
Table 1 (page 320) presents demographic information on the subjects of this study. Three hundred thirty-six subjects were not taking estrogens. The remaining 174 subjects were taking estrogens, the forms of which included esterified estrogens, estradiol, estropipate, estrone and conjugated estrogens. We identified no subjects as taking phytoestrogens for medicinal purposes. Women taking estrogens were better-educated, had a higher monthly family income, were more likely to be European American, and were more likely to be living in the suburban neighborhood than were subjects not taking estrogens.

The women taking estrogens were more likely to have dental health insurance and had received more recent dental care than those not taking estrogens.

Access to and use of health care resources of the two groups of women are compared in Table 2 (page 321). The majority of subjects in both groups had health insurance that covered hospital expenses and physician office visits. The women taking estrogens were more likely to have insurance that covers out-of-hospital prescription drugs. The women taking estrogens also were more likely to have dental health insurance and had received more recent dental care than were those not taking estrogens. In addition, the women taking estrogens held a stronger belief in their ability to personally control their own health, as measured by a three-item personal health control scale.¹⁵

Because the two groups differed in the demographic, sociocultural and health care variables displayed in Tables 1 and 2, we explored the relationship of these variables to the DI, MI and CMI scores using linear multiple regression methods. These analyses uniformly showed that the demographic or health care variables had no statistically significant impact on CMI, DI and MI scores or on the relationship between these scores and estrogen use. This result argues that the potentially confounding demographic, cultural and health care variables did not materially influence the primary findings of this study.

Calculating the CMI score and its component subscale scores involves summing the number of positive symptoms that are disclosed during the examination. Thus, it is conceivable that these summary scores could mask an association between estrogen use and more specific clusters of signs and symptoms. To guard against this possibility, we further subdivided the MI and DI sub-scale scores into symptom cluster scores derived by summing items assessing extraoral masticatory muscle pain (18 items), intraoral masticatory muscle pain (six items), neck muscle pain (12 items), mandibular range of motion (16 items), sounds emanating from the TMJ (four items) and pain on palpation of the TMJ (six items). We repeated multiple-regression analysis as described above to determine the association of estrogen use with each of these symptom cluster scores after controlling relevant sociodemographic, cultural and health care variables. None of these analyses yielded a statistically significant regression coefficient for estrogen use.

Because the subjects in this study did not constitute a clinical sample, we made no attempt to assign a clinical diagnosis. Therefore, we were unable to perform data analysis on the subgroup of subjects who may have met diagnostic criteria for TMD. However, using previous research¹⁶ as a guide, we identified subjects having relatively severe signs and symptoms of a muscle disorder (MI score 0.20) or joint disorder (DI score 0.20), and we compared these subjects with those who had less severe signs and symptoms. Of the entire sample, 23 subjects (4.5 percent) met the criterion for a joint disorder, and 72 (14.1 percent) met the criterion for a muscle disorder. Contrary to expectations, the percentage of subjects meeting these criteria was numerically higher in the group not taking estrogens. Of those taking estrogens, 3.4 percent met the criterion for a joint disorder; 5.1 percent of those not taking estrogens met the criterion. Concerning muscle disorder, 11.5 percent of those taking estrogens met the criterion compared with 15.5 percent of those not taking estrogens. The association between meeting the criteria defined here for a muscle or joint disorder and estrogen status was not statistically significant. We obtained similar results from a multiple logistic regression analysis in which meeting or not meeting these criteria was predicted from estrogen status and covariates representing the sociodemographic, cultural and health care variables displayed in Tables 1 and 2. The analysis of joint disorders yielded a standardized regression parameter of –0.062 with a standard error of 0.298. The odds ratio associated with estrogen use was 0.940 (95 percent confidence interval, or CI, 0.524–1.688). The analysis of muscle disorders yielded a standardized regression parameter of 0.082 with a standard error of 0.178. The odds ratio associated with estrogen use was equal to 1.085 (95 percent CI 0.766–1.537). These findings indicate that after potentially confounding variables (years of education, household income, ethnic group membership, age, availability of dental health insurance, availability of prescription drug insurance, months since last seen by a physician, and months since last seen by a dentist) are controlled for, the odds of meeting an operational definition of more severe vs. less severe joint or muscle disorder are unaffected by estrogen use.

The prevalence and severity of temporomandibular disorder signs and symptoms were nearly identical in women taking and not taking estrogens.

We conducted analyses similar to those described above to compare groups of subjects taking and not taking medications containing progestins or the antiestrogen tamoxifen. Thirty-five subjects (6.9 percent) were taking progestins, and six subjects (1.2 percent) were taking tamoxifen. These analyses revealed no significant differences between the groups taking and not taking these medications.

	DISCUSSION

TOP ABSTRACT SUBJECTS, MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
The results of this study do not support an etiologic or modifying role for exogenous estrogens in the pathogenesis of TMD in a randomly selected community sample of older women. The prevalence and severity of TMD signs and symptoms were nearly identical in women taking and not taking estrogens. This was true both for joint disorders (DI score) and for muscle disorders (MI score). Even when symptom cluster scores representing specific features of the TMD syndrome (such as joint sounds and intraoral muscle tenderness) were considered, we found no evidence suggesting estrogen as a risk factor.

The female subjects included in this study clearly exhibited TMD signs and symptoms at a higher rate than their male counterparts, suggesting they were typical with respect to the often-reported disparity between the sexes regarding this condition. Yet this higher rate of symptom identification could not be explained by their use of exogenous estrogens. The results of this study did not depend on the subjects’ menopausal status. The vast majority of subjects who were taking estrogens were taking replacement estrogens, but a few younger women were receiving estrogen from oral contraceptives. The primary analysis, including all available subjects, produced results that were closely matched by secondary sensitivity analyses excluding women younger than 50 years of age. Raising the cutoff point to 55 and even 60 years of age did not materially affect the outcome. This finding suggests that endogenous estrogens in the ovulating women probably did not confound the outcome.

The findings of this study are at odds with those of LeResche and colleagues,⁸ which demonstrated an association between filling a prescription for an estrogen-containing medication and later being referred for TMD care. The reasons for this discrepancy can only be speculated on. Use of estrogen replacement therapy is not random in postmenopausal women. The decision depends on the nature and severity of the woman’s menopausal symptoms, her medical and family history, and the personal philosophy of patient and physician. Estrogen replacement therapy often is initiated in response to patient behavior. Many physicians will not prescribe replacement therapy unless the patient complains of discomfort or requests estrogen replacement. The same is true for prescription of oral contraceptives; the patient must request the therapy. Thus, availability of health care resources and the patient’s persistence in seeking care will, in part, determine whether she receives estrogen replacement therapy. Our results clearly demonstrate that demographic, socioeconomic, cultural and health care variables distinguish women who are taking estrogens from those who are not. These potentially confounding variables were statistically controlled in this study.

As with estrogen replacement therapy, treatment for TMD likewise may depend on exposure to information about the disorder, the availability of health care resources and the patient’s aggressiveness in seeking care. These circumstances suggest that subject self-selection bias may link estrogen replacement therapy and TMD therapy in clinical samples. LeResche and colleagues⁸ statistically controlled for health care–seeking behavior in their analysis. While the number of recent medical and surgical visits was a strong predictor of TMD referral, the tendency to seek health care did not override the effects of estrogen use.

Subjects in our study were sampled randomly from a nonclinical community population, and the severity of TMD symptoms were quantified by direct physical examination. Our experimental design, by using a random sample, greatly reduces the risk of subject self-selection bias associated with the tendency to seek therapy. Also, we can be fairly confident that subjects who reported taking estrogens and produced a prescription container actually were taking estrogens at the time the oral examination was performed. This cannot be assumed in retrospective studies that rely only on medical records review.

This cross-sectional study, however, was not optimally designed to identify a causal relationship between estrogen use and TMD, and certain limitations should be appreciated. First, a clinical diagnosis of TMD was not made, and we do not know what proportion of our subjects desired treatment at the time of the study or had ever sought treatment in the past. The relationship between estrogens and TMD diagnosis in a clinical sample could be different from the relationship between estrogens and TMD signs and symptoms observed in our community sample. The finding that estrogen use is unrelated to an operational definition of TMD severity argues against this idea.

Another limitation of this study is that the subjects’ menopausal status was not confirmed. It is possible that estrogens play a different role with respect to TMD in ovulating and postmenopausal women. We attempted to address this problem through sensitivity analysis, which demonstrated that subject age, and hence the likelihood that the subject is ovulating, does not appear to affect the relationship between estrogen use and TMD. Finally, this study does not provide information concerning the prescribed dosage of estrogen, current blood levels or the duration of exposure to exogenous estrogens. LeResche and colleagues⁸ demonstrated that the risk of developing TMD increased with a measure of cumulative estrogen exposure. These exposure variables would be expected to play a key role in determining the impact of estrogens on TMD.

	CONCLUSION

TOP ABSTRACT SUBJECTS, MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
In summary, we were unable to confirm an association between use of exogenous estrogens and severity of TMD signs and symptoms in a randomly selected sample of older, primarily post-menopausal women. Therefore, in this nonclinical community sample, estrogen use was not identified as a risk factor for TMD. If, as postulated,⁸ estrogen use raises the risk of referral for TMD care in a clinical population, then we must ask why there appears to be no association between estrogen use and the component signs and symptoms of the TMD syndrome in a randomly selected community sample. Final conclusions regarding any risk of developing TMD associated with estrogen use will have to await the completion of future longitudinal studies using experimental rather than correlational designs.