Article type
Year
Abstract
Introduction: The clinical diagnosis is considered the essential first step in patient management for temporomandibular disorders. However, the diagnostic accuracy of these tests is not clear. This meta-analysis determined the validity of clinical diagnostic tests for the diagnostic subgroup internal derangement using currently available reference standards.
Methods: Strategies used to identify relevant research included computerized bibliographic and conference proceeding database searches from 1983 to 1993, information requests from authors, reference lists. A blinded, independent assessment of articles was done using criteria, human subjects, continuous or binary data, and a diagnostic test comparing a reference standard of diagnosis. Two observers independently assessed each article using methodologic validity criteria. Information regarding data fitting a 2x2 matrix were extracted. The effect estimates of the conditional maximal likelihood estimate of the offs ratio (ORcond), sensitivity (S) and specificity (SP) were calculated for each trial. Zeien's exact test for homogeneity was applied. A common OR (ORCML) using the exact method was calculated.
Results: Seventy-two articles met the selection criteria. Methodologic quality resulted in quadratic weighted kappa from 0.38 to 0.93 for the various categories. MRI (sagittal and coronal views) and arthrotomography (sagittal view) were the more robust reference standards. MRI had a common ORCML of 95.99 (22.68, 675.60) and for arthrotomography it was 82.53 (14.72, 746.40). Two clinical classification schemes resulted in ORcond of higher discriminative power: Schiffman (1989) ORcond of ? (18.44, ?) with S 0.97 (0.82, 1.00) and SP 1.00; DeLaat (1993) ORcond of ? (14.69, ?) with S 0.79 (0.66, 0.88) and SP 1.00.
Discussion: MRI (sagittal and coronal views) and arthrotomography (sagittal views) are the recommended reference standards. The clinical classification schemes by Schiffman and DeLaat have higher discriminative power. The limitation of power in the meta-analyses and use of single small studies to establish diagnostic validity of tests warrant some discretion. Future study designs would be strengthened by adherence to three core methodological elements: the case mix, blinded assessment, and marker and reference tests.
Methods: Strategies used to identify relevant research included computerized bibliographic and conference proceeding database searches from 1983 to 1993, information requests from authors, reference lists. A blinded, independent assessment of articles was done using criteria, human subjects, continuous or binary data, and a diagnostic test comparing a reference standard of diagnosis. Two observers independently assessed each article using methodologic validity criteria. Information regarding data fitting a 2x2 matrix were extracted. The effect estimates of the conditional maximal likelihood estimate of the offs ratio (ORcond), sensitivity (S) and specificity (SP) were calculated for each trial. Zeien's exact test for homogeneity was applied. A common OR (ORCML) using the exact method was calculated.
Results: Seventy-two articles met the selection criteria. Methodologic quality resulted in quadratic weighted kappa from 0.38 to 0.93 for the various categories. MRI (sagittal and coronal views) and arthrotomography (sagittal view) were the more robust reference standards. MRI had a common ORCML of 95.99 (22.68, 675.60) and for arthrotomography it was 82.53 (14.72, 746.40). Two clinical classification schemes resulted in ORcond of higher discriminative power: Schiffman (1989) ORcond of ? (18.44, ?) with S 0.97 (0.82, 1.00) and SP 1.00; DeLaat (1993) ORcond of ? (14.69, ?) with S 0.79 (0.66, 0.88) and SP 1.00.
Discussion: MRI (sagittal and coronal views) and arthrotomography (sagittal views) are the recommended reference standards. The clinical classification schemes by Schiffman and DeLaat have higher discriminative power. The limitation of power in the meta-analyses and use of single small studies to establish diagnostic validity of tests warrant some discretion. Future study designs would be strengthened by adherence to three core methodological elements: the case mix, blinded assessment, and marker and reference tests.