|Home | About | Journals | Submit | Contact Us | Français|
Purpose: To assess the diagnostic validity of clinical tests for temporomandibular internal derangement relative to magnetic resonance imaging (MRI). Methods: MEDLINE and Embase were searched from 1994 through 2009. Independent reviewers conducted study selection; risk of bias was assessed using Quality Assessment of studies of Diagnostic Accuracy included in Systematic reviews (QUADAS); ≥9/14) and data abstraction. Overall quality of evidence was profiled using Grading of Recommendations Assessment, Development, and Evaluation (GRADE). Agreement was measured using quadratic weighted kappa (κw). Positive (+) or negative (−) likelihood ratios (LR) with 95% CIs were calculated and pooled using the DerSimonian–Laird method and a random-effects model when homogeneous (I2≥0.40, Q-test p≤0.10). Results: We selected 8 of 36 studies identified. There is very low quality evidence that deflection (+LR: 6.37 [95% CI, 2.13–19.03]) and crepitation (LR:5.88 [95% CI, 1.95–17.76]) as single tests and crepitation, deflection, pain, and limited mouth opening as a cluster of tests are the most valuable for ruling in internal derangement without reduction (+LR:6.37 [95% CI, 2.13–19.03]), (−LR:0.27 [95% CI, 0.11–0.64]) while the test cluster click, deviation, and pain rules out internal derangement with reduction (−LR: 0.09 [95% CI, 0.01–0.72]). No single test or cluster of tests was conclusive and of significant value for ruling in internal derangement with reduction. Conclusions: Findings of this review will assist clinicians in deciding which diagnostic tests to use when internal derangement is suspected. The literature search revealed a lack of high-quality studies; further research with adequate description of patient populations, blinded assessments, and both sagittal and coronal MRI planes is therefore recommended.
Objectif : Évaluer la validité du diagnostic résultant de tests cliniques en imagerie de résonance magnétique (IRM) pour le dépistage du dérangement interne de l'articulation temporo-mandibulaire (ATM). Méthode : Une recherche a été effectuée dans les bases de données MEDLINE et Embase pour les années 1994 à 2009. Des examinateurs indépendants ont procédé au choix des études répertoriées ; les risques de biais ont été évalués à l'aide de l'échelle QUADAS (Quality Assessment of studies of Diagnostic Accuracy included in Systematic reviews), visant à évaluer la qualité des études sur la fiabilité diagnostique incluses dans les revues systématiques (le pointage obtenu a été ≥9/14) et à l'aide d'abstraction des données. Le profil de la qualité globale des preuves a été établi avec l'échelle GRADE (Grading of Recommendations Assessment, Development, and Evaluation). Le degré d'accord a été mesuré à l'aide du coefficient quadratique kappa pondéré (κw). Les rapports de vraisemblance (RV) positifs (+) ou négatifs (−) avec intervalles de confiance de 95 % ont été calculés et groupés suivant la méthode de DerSimonian–Laird et à l'aide d'un modèle à effets aléatoires lorsque les données étaient homogènes (I2≥0,40, test-Q p≤0,10). Résultats : Nous avons retenu 8 des 36 études identifiées. Les preuves voulant que les tests individuels de déviation (RV+ : 6,37 [95 % IC : 2,13–19,03]) et de crépitation ([RV : 5,88 [95 % IC : 1.95–17.76]) soient fiables étaient de très faible qualité, alors qu'un ensemble de tests regroupant crépitation, déviation, douleur et ouverture limitée de la bouche sont les plus valables au moment de déterminer s'il y a effectivement dérangement interne sans réduction (RV+ : 6,37 [95 % IC : 2,13–19,03]), (RV− : 0,27 (95 % IC : 0,11–0,64]), alors que les tests regroupés – claquement, déviation et douleur – déterminent habituellement s'il y a dérangement interne avec réduction (RV− : 0,09 [95 % IC : 0,01–0,72]). Aucun test unique ni aucun groupe de tests ne semblent avoir de valeur considérable ou significative ni concluante pour déterminer avec précision un dérangement interne avec réduction. Conclusions : Les conclusions de cet examen aideront les cliniciens à décider quels tests diagnostics utiliser lorsqu'ils soupçonnent un dérangement interne. La recherche documentaire a révélé un manque d'études de grande qualité ; des recherches supplémentaires comportant une description adéquate des populations de patients, des évaluations à l'aveugle et des plans sagittaux et coronaires de l'IRM sont donc recommandés.
Temporomandibular joint (TMJ) internal derangement is one of the most common forms of temporomandibular disorders (TMD). The most prevalent signs and symptoms associated with these disorders are tenderness of the masticatory muscles, pain, joint noise, and decreased range of jaw movement.2 Although these clinical manifestations are common, only a small proportion of those affected (as little as 5%3 and 7%4) seek help.
The natural history of TMD has received minimal study. It is thought that signs and symptoms are transient, self-limiting, and frequently without serious long-term effects.3 Others have suggested that minor dysfunction may progress to more serious joint disease.5,6
The American Academy of Orofacial Pain has estimated that between 40% and 75% of the U.S. population displays at least one sign of TMD, while 33% report at least one symptom.6 In addition, the 2002 U.S. National Health Institute Survey (NHIS) of 31,000 individuals found an overall prevalence of TMJ and muscle-disorder-type pain of 4.6% (6.3% in women, 2.8% in men).7 Among the general population, TMD is classified as a condition of young and middle-aged adults rather than of children or the elderly.8 It was estimated in 1999 that 5.3 million Americans seek treatment for TMD over the course of a 6- to 12-month period, resulting in an estimated $2 billion in direct costs alone.9
Numerous imaging modalities are currently in use to assess TMJ disorders, but a gold standard has not been established to identify the “true” clinical diagnosis.2 Arthrography has been reported to have the best diagnostic outcomes for identifying internal derangement (ID);10 arthrography is invasive, however, and has poor inter-observer performance relative to magnetic resonance imaging (MRI). Our study used MRI as the reference standard, as it has been found to be the most accurate modality in detecting disk positions and can visualize bilateral joints simultaneously.1,11
Clinical diagnosis is the first step in management of TMD, yet information about the validity of TMD tests is very limited. Tests to identify TMD can be either single or clustered; a single test evaluates individual symptoms, such as pain or joint noise, whereas a cluster of tests can assess the cumulative effects of a variety of findings (e.g., TMJ range of motion, pain, and joint noise) to determine a clinical diagnosis. At present, numerous clinical diagnostic tests are administered in clinical settings. The combined ability of these tests to identify ID of the TMJ, however, has not been determined.
Research comparing clinical diagnostic tests with MRI findings to distinguish among TMD types has demonstrated inconsistent results.11 The purpose of this systematic review of prospective cohort and case-control studies is to determine the diagnostic validity of clinical tests for ID relative to MRI findings. ID includes three categories: internal derangement of any kind (ID), ID with reduction (IDR), and ID without reduction (IDnoR).
Studies were considered for the review if they met the following criteria: case-control or prospective cohort study; participants aged ≥14 years with TMD, ID, IDR, or IDnoR; study using tests easily performed in a clinical setting with minimal equipment, either individual (single) or a group of tests (clustered); MRI images including sagittal and coronal planes used.
We searched MEDLINE and Embase (with English language restrictions) for medical, specialist, and allied health literature from January 1, 1994, to October 1, 2009. MeSH subject headings and keywords used were temporomandibular joint disorders, validity, validation, diagnosis, sensitivity and specificity, predictive value of tests, and magnetic resonance imagery. The search was limited to studies involving humans. Our search strategies are outlined in Appendix A.
Three authors independently reviewed the articles and made selections based on the relevance of the title and abstract to the research topic. Calibration was based on the first 50 citation postings for title and abstract screening; for full-text screening, calibration was performed on a training set before article selection. Screening criteria were applied as outlined above.
Following calibration, the QUADAS quality assessment tool12 was used by two independent reviewers to assess internal validity (the degree to which the diagnostic tests truly evaluate ID and not other variables; see Table Table1).1). Consensus was reached through discussion. Articles scoring ≥9 on the 14-item validity tool were advanced to data abstraction.
To determine the external validity of the diagnostic tests (the degree to which the results are generalizable to other populations or environmental settings), the GRADE system for grading the body of evidence was used. This system, first developed in 2004 by an informal working group with the goal of surmounting the shortcomings of the current grading systems,13,14 takes into account
in determining the quality of evidence for each important outcome.13,15 Based on this information, GRADE classifies the level of evidence into one of four categories: high, moderate, low, or very low.13,14 As of 2009, more than 32 medical societies, health regulatory bodies, and health-related branches of government had adopted the GRADE system.16 It should be noted, however, that a recent review of the GRADE system noted that it has not yet been proven to be reliable or valid, and identified the potential for bias.16 The GRADE score (reported in Table Table2)2) was determined by two independent reviewers, and consensus was reached through discussion.
Three independent reviewers extracted data from eight validated articles. Data were extracted based on MRI type, clinical test, referral pattern, study design, and sample population (prevalence) (see Table Table33).
Our data analysis included the following five steps: (1) inter-rater reliability testing for study selection and methodological quality assessment; (2) calculation of test accuracy measures such as sensitivity, specificity, and likelihood ratios using Meta-DiSc24; (3) testing of heterogeneity; (4) pooling of data; and (5) sub-group analysis.
For article selection and methodological quality assessment, inter-rater reliability between evaluators was calculated using the quadratic weighted kappa statistic (Cicchetti weights [κw with standard deviation (SD)]).25 For data extraction, raw data were retrieved for 2×2 table construction. Pre-test probability was calculated. The data sets were categorized based on MRI view, clinical test type, and pathological condition. Analysis was based on ID and sub-group pathology of IDR and IDnoR.
Meta-DiSc24 was used to generate sensitivity, specificity, and likelihood ratios (LR). A 95% confidence interval (CI) was applied. Accuracy can be expressed in terms of sensitivity (the proportion of positive test findings among people with the disorder) and specificity (the proportion of negative test results among people without the disorder) or by LRs; LRs are the most clinically helpful in reporting diagnostic accuracy of clinical tests. The LR incorporates the sensitivity and specificity of a test into a single measure and is independent of the prevalence of the disorder within a given population. A positive LR (+LR) indicates how much more likely a positive test result is in people who have the disorder than in those who do not; the ideal test for ruling in a disorder is the one with the largest +LR. A negative LR (−LR) indicates how much more likely a negative test result is in people without the disorder than in people with it; the best test for ruling out a disorder is the one with the smallest −LR. Values in Table Table44 show the clinical application of LRs in assessing a shift in the probability of the disorder's being present; that is, a +LR>10 indicates a large and often conclusive probability that the condition is present, while a −LR<0.10 suggests a large and often conclusive probability that the condition is not present.26 A +LR or −LR of 1 means that a positive or negative result, respectively, is equally probable in a participant with and a participant without the disorder. A CI including an LR of 1 should therefore be interpreted with caution. The DerSimonian–Laird method was chosen for estimating the pooled likelihood ratios (LRp), using a random-effects model.27
When studies were deemed clinically similar enough for pooling of results, the heterogeneity of studies was investigated first graphically and then statistically. Cochran's Q-test was calculated, and a p-value ≤0.10 was considered to indicate statistical heterogeneity.28,29 Inconsistency was measured using I2, a method of quantifying heterogeneity in a meta-analysis.30 An I2 of 0% for a trial indicates that all variability in effect estimates is due to sampling error within trials and none is due to heterogeneity; an I2 of 40% indicates that 40% of variability between the trials can be attributed to study variation, and is considered mildly to moderately inconsistent.
Sub-group analysis was planned a priori for three factors: (1) methodological quality (poorer trial quality indicated by a QUADAS score ≤9/14); (2) pre-test probability levels, that is, the type of population, from general to tertiary care (usually low in general care, intermediate in secondary care, and high in tertiary care); and (3) publication bias. This analysis was not conducted, however, because we did not have sufficient data for a meta-regression to be useful.31
Our database search identified 4,229 citation postings between January 1994 and October 2009 (see Figure Figure1).1). Based on relevant abstracts and titles, 36 articles were retrieved for full-text screening. The final results were based on 8 primary studies (κw 0.66, SD=0.24). Studies were excluded for reasons relating to sample population (n=4), clinical testing (n=5), MRI (n=3), 2×2 table construction (n=14), disorder type (n=1), and study design (n=7).
A single study19 examined ID, IDR, and IDnoR from the sagittal plane only, using a cluster of clinical tests (see Table 2). The other seven studies examined the sagittal and coronal views.1,17,18,20–23 Of the eight selected articles,1,17–23 seven examined ID1,17–22 (cluster n=3, single n=3, both n=1) and seven examined both IDR and IDnoR1,17–21,23 (cluster n=4, single n=2, both n=1). All measures within these clusters must be scored positive for the diagnosis to be made, although reporting of the scoring method was not consistently detailed or transparent. One study was based on primary referral and six on tertiary referral; the remaining study was unclear on this point (see Table 3).
All articles selected were of high quality (≥9/14 QUADAS score; see Table 1). Each study consistently administered both the clinical test(s) and MRI to all participants. In addition, the studies explained both reference and clinical test(s) in depth, which ensures reproducibility. Two factors determined poorer quality. First, five studies (62.5%) did not specify the duration between MRI and administration of clinical test(s), which allows for the possibility that natural history changed the condition. Second, blinding of evaluators to the MRI or clinical test(s) was not consistently reported: three studies did not report any blinding, and three studies reported single blinding. Lack of blinding can bias the evaluator's interpretations of the data.
We identified four trials17–20 that explored the presence or absence of a click through palpation for various types of ID. Although these trials appear to have had the same type of clinical setting, used the same reference tests and clinical tests, and had similar population characteristics, they were heterogeneous in terms of pre-test probability. This suggests that even though the participants were from a tertiary-care population they may have varied along the spectrum of the disorder. We therefore did not pool the results.
There is very low to moderate-quality evidence that the presence of any click is of rare importance in identifying ID (4 trials,17–20 369 participants), IDR, or IDnoR (3 trials,17–19 369 participants). Similar findings exist for atypical reciprocal click (ARC, opening click separated from closing click by less than 5 mm) and single click (SC, opening or closing click accompanied by excursion and protrusion click).
There is low-quality evidence (1 trial,17 39 participants) suggesting that a reciprocal click (RC) has a small and sometimes important probability in ruling in or ruling out IDR (+LR: 3.24 [95% CI, 0.90–11.67]; −LR: 0.50 [95% CI, 0.31–0.80]), IDnoR (+LR: 2.14 [95% CI, 0.54–8.51]), and ID (+LR: 2.89 [95% CI, 0.81–10.39]). The −LRs have very little importance for IDnoR and ID, and are therefore not reported.
We identified five trials1,19–22 that explored the presence of pain on palpation, functioning, and opening in various types of ID. Two studies1,21 had similar population characteristics and pre-test probabilities and were not statistically heterogeneous (for IDR, I2=0%; p=0.34–0.38; for IDnoR, I2=0%; p=0.49–0.68; for ID, I2=0–48%; p=0.14–0.39). Pooling of their results for further analysis was therefore possible; Figures Figures22–4 illustrate the results. The remaining three studies19,20,22 were heterogeneous because of the differences in how their diagnostic test(s) were defined, which prevented any further meta-analysis. One article20 did not present sufficient data for us to summarize its findings.
There is moderate-quality evidence (2 trials,1,21 172 participants) suggesting that pain on palpation, function, or opening has a very small and rarely important probability of identifying the presence of IDR (+LRp: 1.44 [95% CI, 1.02–2.04]; −LRp: 0.80 [95% CI, 0.63–−1.00]). The same 2 trials1,21 (172 participants) offer moderate-quality evidence suggesting that pain on palpation, function, or opening is of small importance and can sometimes assist in identifying IDnoR (+LRp: 2.11 [95% CI, 1.58–2.82]; −LRp: 0.47 [95% CI, 0.36–0.62]). There is also moderate-quality evidence from three trials1,21,22 (366 participants) that pain on palpation, function, or opening is of very small importance and can rarely assist in identifying patients with any ID (+LRp: 1.65 [95% CI, 1.36–2.01]; −LRp: 0.66 [95% CI, 0.54–0.80]).
We identified three trials17,19,23 that investigated the use of a cluster of tests including ROM, muscle palpation, joint palpation, and joint sounds for ID. All tests in this cluster need to be positive to identify ID. Although the clinical tests performed were similar, heterogeneity in diagnostic criteria, reference standard (MRI of sagittal vs. coronal view), referral source, and pre-test probability precluded pooling of results, with the exception of IDR and IDnoR findings for two of the articles.17,23
There is low-quality evidence (2 trials,17,23 281 participants) to suggest that this cluster of tests is a small and sometimes important indicator in identifying patients with IDR (+LRp: 2.97 [95% CI, 2.36–3.74]; −LRp: 0.42 [95% CI, 0.28–0.61]). Although the data showed some variation in pre-test probability, they were deemed clinically similar enough to pool and were not statistically heterogeneous (I2=0–34%; p=0.21–0.74). In addition, there is low-quality evidence (2 trials,17,23 281 participants) suggesting that the same cluster of tests is a small and sometimes important indicator for ruling in IDnoR in the presence of a positive cluster of tests (+LRp: 4.13 [95% CI, 3.02–5.64]) and has a very small and rarely important probability of ruling out IDnoR in the presence of a negative cluster of tests (−LRp: 0.66 [95% CI, 0.37–1.16]). Although these trials were statistically heterogeneous (I2=0–90%; p=0.00–0.89), we decided to pool them because they were deemed to be clinically similar.
One trial19 with 40 participants examined the effectiveness of clusters of tests—including (a) click, deviation, and pain, and (b) crepitation, deflection, pain, and limited mouth opening—in determining the presence of ID. This trial yielded very low quality evidence that test cluster (a) presence of click, deviation, and pain is often conclusive in ruling out IDR in the presence of a negative cluster of tests (−LR: 0.09 [95% CI, 0.01–0.72]). In addition, there was very low quality evidence that test cluster (b) crepitation, deflection, pain, and limited mouth opening is a moderate and usually important indicator for ruling in IDnoR in the presence of a positive test (+LR: 6.37 [95% CI, 2.13–19.03]) and a small and sometimes important indicator for ruling out IDnoR in the presence of a negative test (−LR: 0.27 [95% CI, 0.11–0.64]). A cluster of tests appears to be better at ruling a condition in or out than a single test.
Furthermore, there was very low quality evidence from this same trial suggesting that another test cluster (c) crepitation, deflection, pain, and limited mouth opening has a small and sometimes important probability of ruling in ID of any kind in the presence of a positive cluster of tests [(+LR: 2.27 [95% CI, 0.72–7.15]) and a very small and rarely important probability of ruling out ID of any kind in the presence of a negative cluster of tests (−LR: 0.83 [95% CI, 0.67–1.03]).
Heterogeneity precluded meta-analysis of trials (2 trials,19,20 113 participants) indicating crepitation as a test in identifying participants suffering from ID. One trial19 with 40 participants yielded very low quality evidence suggesting that the sign of crepitation has a moderate and usually important probability that IDnoR is present (+LR: 5.88 [95% CI, 1.95–17.76]). In addition, the absence of crepitation has a small and sometimes important probability of ruling out IDnoR (−LR: 0.33 [95% CI, 0.16–0.71]) in the presence of a negative test finding.
There was very low quality evidence (1 trial,19 40 participants) that the sign of crepitation is a small and sometimes important indicator for ruling in ID (+LR: 2.42 [95% CI, 0.78–7.57]). Other clinical tests were suggestive of ID. The presentation of deviation is a very small and rarely important indicator of IDR (+LR: 1.00 [95% CI, 0.86–1.16]), as is the absence of deviation (−LR: 0.99 [95% CI, 0.18–5.49]). LR values for deviation as a predictor for or against IDnoR or for ID were not significant.
Deflection as a positive indicator for IDnoR is considered a moderate and usually important sign (+LR: 6.37 [95% CI, 2.13–19.03]), and deflection as a negative indicator of IDnoR is considered a small and sometimes important sign (−LR: 0.27 [95% CI, 0.11–0.64]). The presence of deflection is a small and sometimes important sign to rule in ID of any kind (+LR: 2.58 [95% CI, 0.83–8.00]).
The goal of this review was to assess the diagnostic validity of clinical tests for TMD classified as IDR or IDnoR relative to MRI. The review provides no evidence to support any one clinical test as a significant and conclusive predictor of the presence or absence of ID relative to MRI; however, there is evidence that certain tests may be of some relevance in helping to diagnose TMD.
In reviewing the extracted data, we decided to base our recommendations on the LRs of the diagnostic tests rather than on their specificity and sensitivity. Although specificity and sensitivity provide useful information, they cannot be used to quantify the shift in probability of the condition given a certain test result.32 By contrast, the LR combines the information contained in sensitivity and specificity with the pre-test probabilities to determine the shift in probability based on the results of the diagnostic test.
Although there was no high-quality evidence showing strong and conclusive probability of clinical tests' diagnosing the presence or absence of IDR, IDnoR, or ID, there are some tests with moderate and others with small diagnostic properties, as follows (see Table Table55):
Our literature search did not find any other current systematic reviews to which we could compare our results. One unpublished thesis2 reviewed clinical tests compared to several other reference standards. The author determined that one clinical classification scheme (IDR: RC, no coarse crepitus, open ≥35 mm; IDnoR: history of movement limitation, no RC, no coarse crepitus, maximum opening ≤35 mm, passive opening stretch <40 mm, contralateral movement <7 mm, no SD) had high discriminative power (n=1, pre-test probability: 0.85; sensitivity: 0.97 [95% CI, 0.82–1.00]; specificity: 1.00 [95% CI, unable to calculate]); however, the reference standard was arthrotomography, and the methodological quality of the study was low, not unlike the clusters determined in this review.
Despite efforts to limit bias and ensure good methodological quality, this study has several potential limitations. A comprehensive search for relevant literature was conducted using two online databases (MEDLINE and Embase) to retrieve published literature; our methods did not include a search of the grey literature (e.g., bibliographical lists, unpublished data), which may have introduced selection bias as this potentially relevant information may not have been found. In addition, publication bias may have influenced our results, as only English-language studies were reviewed.33 Although sensitivity analysis for publication bias was planned, heterogeneity of studies precluded pooling of results across more than three trials, and funnel plots were therefore not constructed.
Meta-analysis of the data was limited by several issues. Although study participants were mostly examined at tertiary centres, a few studies included patients from primary-care facilities. In addition, clinical tests of the patient population often had dissimilar pre-test probabilities, despite the participants' being from the same referral source (either primary or tertiary). This would lead us to believe that patients were presenting with different stages of the disease. Therefore, despite the similarity of study characteristics, we did not pool data from studies whose pre-test probabilities were substantially different. Large differences in pre-test probabilities also raise concerns about the presence of spectrum bias, which occurs when diagnostic testing is done on individuals who are not representative of the population to which the test is usually applied in practice.32 Review of the included articles revealed, however, that all were prospective cohort studies with consecutive groups of participants from a clinical population, a factor essential to minimizing spectrum bias;32 in addition, all studies included symptomatic patients only. The likelihood of spectrum bias was therefore determined to be minimal.
Another issue was diagnostic test definition, which made meta-analysis sometimes impossible. The test identified as “pain” was variously defined as pain on palpation, function, and opening; pain on palpation; or even self-reported ear or TMJ pain. “Click” was also assessed by different techniques: an audible click, a palpated click, or a click evaluated with a stethoscope. In addition, clusters of tests did not always refer to the same tests. One study,21 for example, grouped different tests after a mathematical evaluation of the LR. Other authors followed the RDC/TMD (Research Diagnostic Criteria for Temporomandibular Disorders) protocol.34
MRI has been shown to be a reliable and valid method for diagnosing ID of the TMJ.35 Studies that satisfied our inclusion criteria all used MRI as the gold standard. However, one study19 used only sagittal images for diagnosis of ID, whereas the other seven studies used both sagittal and coronal images. It has been recommended that both sagittal and coronal images be obtained for optimal MRI assessment of the TMJ.36 Thus, conclusions drawn from the findings of the aforementioned article19 should be implemented with caution in clinical practice. Although this study did suggest that cluster, crepitation, and deflection could have important predictability in determining ID, methodological quality was deemed to be very low for several reasons, including unclear blinding procedures, small sample size, use of sagittal MRI only, and unclear time period between index and reference test.
In designing future diagnostic studies, researchers should focus special consideration on five critical quality domains to avoid systematic error: adequate description of patient population; blinded assessment; a clear definition of the diagnostic test; clear reporting of the duration between clinical and reference tests; and use of both sagittal and coronal MRI views as the reference standard, to ensure higher accuracy and consistency across reference tests. To reduce the risk of random error from imprecision, estimates should be based on an adequate sample size.
Our review indicates that there is a lack of high-quality research examining the accuracy of clinical tests in determining ID. Furthermore, the majority of the clinical tests studied were of little importance in determining the presence or absence of ID. Although some tests display a higher LR, and can therefore be considered to have moderate and usually important probability in determining the presence of ID (and one of significant and large importance in ruling out IDR), the quality of evidence was very low. In other words, the evidence for the test that was of usual or large importance to the diagnosis of ID comes from a single small study with moderate or high risk of bias. However, taking into consideration lower-quality evidence and other limitations, as outlined above, we can draw several conclusions. There is no single or cluster of tests of moderate or high quality that has more than small importance in determining IDR. Contrary to clinical reports, click does not demonstrate significant importance in ruling in IDR. The most important findings in ruling in IDR, according to our review, are (a) a test cluster including ROM, muscle–joint palpation, and joint sounds (1 meta-analysis); and (b) reciprocal click as a single test (1 study). Despite their small importance, there is moderate-quality evidence to suggest that both are effective in ruling in IDR.
Deflection and crepitation are the most valuable single tests (1 study), and the test cluster including crepitation, deflection, pain, and limited mouth opening (1 study) is most valuable to rule in IDnoR, while the test cluster click, deviation, and pain helps to rule out IDnoR in the absence of a positive test. However, the evidence to support this statement is of very low quality (1 study) and needs replication. Conversely, there is moderate-quality evidence (1 meta-analysis), with low clinical importance, of pain in TMJ on palpation, opening, and function, and low-quality evidence (1 meta-analysis) that a test cluster of ROM, muscle–joint palpation, and joint sound is helpful in ruling in IDnoR.
Internal derangement (ID) is one of the most common causes of TMD. As a result, the diagnostic validity of tests such as pain on palpation, click, deflection, limited mouth opening, and crepitation to identify ID has been studied in the literature. Clusters of tests have also been examined for their diagnostic accuracy. However, methodological quality concerns have been identified that call into question the clinical effectiveness of implementing these tests in practice. Both click and pain have been suggested as good indicators of ID, but no systematic review of the diagnostic accuracy of clinical tests has been published.
This systematic review summarizes the results of the current literature on the diagnostic validity of tests to identify ID in patients with TMD. Most importantly, careful attention was paid to including studies of sound methodological quality. In addition, the GRADE system was used to assess external validity and determine the strength of recommendations for the diagnostic tests and strategies. The quality of evidence varied from very low to moderate; this review reflects low-quality evidence that deflection and crepitation are the most valuable single tests to determine IDnoR and that crepitation, deflection, pain, and limited mouth opening is the most valuable cluster of tests to determine IDnoR. A negative result for the test cluster click, deviation, and pain is often conclusive that the condition IDR is absent. Despite moderate-quality evidence, no single test (including click) or cluster of tests was of more than small importance in identifying the presence of IDR.
((((tmj)) AND ((“diagnosis”[Subheading] OR “diagnosis”[All Fields] OR “diagnosis”[MeSH Terms] OR diagnos*))) AND ((mri))) AND ((sensitivity specificity OR validity OR predictive value of tests OR validation))
Physiotherapy Canada 2012; 64(2);116–134; doi:10.3138/ptc.2010-54