Diagnostic outcome and observer performance in ... - BIR Publications

4 downloads 0 Views 1MB Size Report
degenerative lesions of the temporomandibular joint was studied. Thirty tomograms depicting findings such as cyst, erosion, osteophyte and sclerosis and 30 ...
Diagnostic outcome and observer performance in sagittal tomography of the temporomandibular joint W. Cholitgol, A. Petersson", M. Roblin·, K. Tanimoto t and S. !kennan+ Department of Oral Radiology, Faculty of Dentistry, Chiangmai University, Thailand, "Department of Oral Radiology, School of Dentistry, Lund University, Sweden, "Department of Oral Radiology, School of Dentistry, Hiroshima University, Japan and 'Department of Stomatognathic Physiology, School of Dentistry, Lund University, Sweden

Received 29 March 1989 and in final form 18 September 1989 The diagnostic outcome and observer performance of sagittal tomography in detecting degenerative lesions of the temporomandibular joint was studied. Thirty tomograms depicting findings such as cyst, erosion, osteophyte and sclerosis and 30 with a normal appearance were selected. The joint status was verified histologically. Four observers evaluated the tomograms with the aid of reference tomograms, independent of the microscopic examination. The diagnostic accuracy was high (80-87%). The sensitivity was between 67% and 90% and the specificity between 73% and 93%. An osteophyte was, with few exceptions, a true finding whereas sclerosis most frequently was false. About two-thirds of the reports of cyst and erosion were found to be true. The interobserver overall agreement rates for any two observers varied between 68% and 90%, for three observers between 65% and 82%, and for all four observers was 63%. The Kappa value for any two observers was 0.40-0.80, indicating fair to substantial agreement. The intra-observer agreement was only somewhat higher than the interobserver. This study shows that, with the aid of reference tomograms, high diagnostic accuracy and observer agreement can be achieved in sagittal tomography. Keywords: Osteoarthritis; temporomandibular joint; temporomandibular joint diseases; tomography, X-ray

Radiographic examination of the temporomandibular joint (TMJ) is an integral stage in the diagnosis and treatment planning of disorders of the stomatognathic system and may also be required for the evaluation of a therapeutic regimen. Sagittal tomography has been shown in clinical studies to yield more diagnostic information than other radiographic methods':", and in autopsy studiesv" to be superior to, for example, lateral oblique transcranial radiography in detecting structural changes. Its diagnostic outcome, that is its accuracy, sensitivity and specificity, is influenced by the severity, extent and location of the degenerative changes'' Of equal concern is the standard of observer performance. In a clinical study of patients with rheumatoid arthritis'" observer agreement was higher with tomography compared with panoramic radiography. However, this technique has not been evaluated against a gold standard determined from knowledge of the 'true' status of the TMJ, nor have the uncertainties due to the variable skills of the participating observers been taken into consideration. This study was therefore undertaken to evaluate sagittal tomography for the diagnostic outcome in detecting microscopically verified bony changes and second the observer performance.

Materials and methods The study was based on a comparison of tomographic and histological examinations of human TMJ autopsy

specimens from elderly individuals (ranging in age from 60 to 88 years) of both sexes who had donated their bodies to research. An oral radiologist, not included among the observers, selected 60 high-quality tornograms evenly distributed between the lateral, central and medial parts of the joint. Thirty tomograms were normal (Figure la), while the remainder had pathological features evenly distributed between the condyle and temporal component (Figures 2a, 3a, 4a). All the corresponding histological sections were examined and those for the 30 tomograms with radiological findings had, in all, 18 cysts and erosions, 15 osteophytes and 10 examples of sclerosis. Out of the 30 sections, four revealed three, five two and the remaining 21 revealed one pathological feature. Thus, four tomograms showed simultaneously erosion (or cyst), osteophyte and sclerosis and five tomograms an erosion together with an osteophyte.

Radiographic technique The specimens were taken en bloc, approximately 6 X 6 X 6 em" in size and attached to a positioning device for tomography. The joint was orientated so that the long axis of the condyle was perpendicular to the tomographic plane as described by Omnell and PeterssonCorrected sagittal tomography was performed with a Polytome U unit (MassiotlPhilips, Paris, France) with a Bi 125/30/50R X-ray tube (Siemens, Erlangen, FRG) with a nominal focus size of

© 1990 IADMFR 02SQ-832 x 90/01 OOOI-{)6

Dentomaxillofac. Radiol., 1990, Vol. 19, February

1

Sagittal tomography of the TMJ: W. Cholitgul et al.

Figure 1 Sagittal tomogram (left) showing a temporomandibular joint which was judged by all four observers to be normal. Photomicrograph

(right) of the corresponding area confirming the normal morphology

Figure 2 Sagittal tomogram (left) showing an erosion (arrow) in the condyle about which all four observers were unanimous. The corres-

ponding photomicrograph (right) shows the bone resorption (arrow) and also a bony exophyte (arrowhead). Superior to the erosion there is a perforation of the posterior disc attachment: the malformed disc is anteriorly displaced

O.6xO.6 mrn". A hypocycloidal movement was used with a tomographic angle of 48° without a grid. The diameter of the circular collimator was 1.5 em, resulting in a field size diameter of 6.5 em at the film plane. Exposure data were 60kVp, 20-33mA and 6s. A multi-film cassette with five pairs of CaW0 4 screens (the last pair of screens were Saphir and the others Rubin, Siemens-EIema, Sweden) and five films (Ilford Rapid R, Ilford, UK), was used. Two cassettes were required so as to cover the full mediolateral extent of the joint. Radiographic assessment Four observers (three oral radiologists from different countries and one stomatognathic physiologist) evaluated the tomograms twice with a period of 1 month between the readings, independent of the histological results. The observers were informed that there was a 2 Dentomaxillofac. Radiol., 1990, Vol. 19, February

50% a priori probability that the joints had structural abnormalities, but not of the frequency of the single findings, nor that nine tomograms depicted more than one. Both the condylar and temporal components were assessed as normal (Figure 1) or with bony abnormalities which were characterized as the following: Cyst - a well-defined, local area of bone rarefaction underneath an intact cortical outline of the joint surface. Erosion - a local area with decreased density of the cortical joint surface and adjacent subcortical bone (Figure 2). Osteophyte - a marginal bony outgrowth on the anterior part of the condyle (Figure 3). Only examples of extensive osteophytes were selected. Sclerosis - a local area with increased density of the cortical bony joint surface extending into the subcortical bone (Figure 4).

Sagittal tomography of the TMJ: W. Cholitgul et al.

Figure 3 Sagittal tomogram (left) showing an extensive osteophyte anteriorly on the condyle on which all observers agreed. Photomicrograph (right) of the corresponding area

., ~

t'

W



~

,

Figure 4 Sagittal tomogram (left) showing sclerosis in the temporal component (arrow). Only one of the observers reported this as a finding.

Photomicrograph (right) of the corresponding area with a localized area of thickened trabeculae extending into the subcortical bone

A series of tomograms of the lateral, central and medial parts of joints depicting both normal and these abnormal features, which had been verified microscopically, were selected as reference images for use by the four observers during their assessments. Histological examination

The specimens were demineralized in 0.5 M ethylenediaminetetra-acetate for 16-20 weeks and then embedded in paraffin. Microtome sections, 6ILm thick, were cut sagittally and stained with Mayer's haemalun-eosin solution 10. Both the soft and hard tissues in the region corresponding to the tomogram were evaluated. An example of normal morphology is shown in Figure 1, an erosion in Figure 2, an osteophyte in Figure 3 and an example of sclerosis in Figure 4.

Mathematical methods

Calculations were based on an overall diagnosis, that is joints with or without any radiological finding, as well as on the presence or absence of the individual findings of cyst, erosion, osteophyte or sclerosis. The diagnostic accuracy is the percentage of correct diagnoses. The sensitivity and the specificity were defined as detailed below!'. Sensitivity is the proportion of joints where a microscopic change can be detected on the corresponding radiograph and is expressed as: true positives true positives + false negatives Specificity is the proportion of joints with normal histology and no abnormal radiological findings, and is Dentomaxillofac. Radiol., 1990, Vol. 19, February 3

Sagittal tomography of the TMJ: W. Cholitgul et al. actual number reported. The false negatives consisted either of an erosion or, more frequently, of sclerosis in the articular eminence.

expressed as: true negatives true negatives + false positives

Observer performance

The accuracy, sensitivity, specificity and interobserver performance were calculated from the second readings. The interobserver and intra-observer performance was expressed as the overall agreement; for the former, the total number of joints on which the observers agreed and, for the latter, the total number of joints on which one observer agreed at both the first and second readings. The observer performance was also expressed as the Kappa index, calculated according to Cohen 12. This index takes into account the contribution to agreement of chance and is expressed as: Po-Pc I-Pc where Po = the proportion of radiographs on which the judges agreed and Pc = the proportion of units for which agreement is expected by chance.

Results Accuracy, sensitivity and specificity

Table I presents the number of true and false diagnoses made by the four observers. Their overall diagnostic accuracy was 80%, 80%, 82% and 87%. Observer D who had the lowest sensitivity (67%) also had the highest specificity (93%), while the converse was true for Observer B. All nine joints with more than one finding were correctly assessed. The number of radiological specific findings reported by each observer is presented in Table II. An osteophyte was, with few exceptions, a true positive finding in contrast to sclerosis which most frequently was a false positive. Erosions and cysts were found to be true positive in about two-thirds of cases irrespective of the Table I Overall number of true and false diagnoses and sensitivity and specificity of sagittal tomography of 60 temporomandibular joints

Observer A B C D Mean

True positive

False negative

False positive

True negative

24 27 27 20

6 3 3 10

6 8 5

24 22 25 28

2

Sensi- Specitivity ficity ('Yo) ('Yo) 80 73 83 93 82

80 90 90 67 82

Table n Number of true-positive and false-positive individual radiological findings reported by four observers for sagittal tomography of 60 temporomandibular joints

Cyst and erosion' Osteophyte'

Sclerosis'

True False True False True False Observer positive positive positive positive positive positive A B C D

13 13 16 8

6 6 8 3

15 15 15 14

1

2 1

o

7 4 3 1

7 9 7 2

Number of microscopical findings: ·18; tIS; :1:10.

4 Dentomaxillofac. Radiol., 1990, Vol. 19, February

The interobserver agreement rates are presented in Table III. Overall they varied between 68% and 90% for each pair of observers and the Kappa index between 0.40 and 0.80. There was little difference when three observers were compared and, for all four observers, the concordance was 63%. The intra-observer agreement rates (Table IV) were only somewhat higher than the interobserver: overall, they varied between 78% and 90% with a Kappa index of between 0.58 and 0.79. Table III Interobserver performance in sagittal tomography of 60 temporomandibular joints and the number of positive, true-positive, negative and true-negative diagnoses

Diagnoses Overall agreement Kappa True True index Positive positive Negative negative Observers (%) AlB A/C AID B/C BID CID AIBID AlBIC A/C/O B/CID AIBICID

85 90 77 88 68 73 65 82 70 65 63

0.70 0.80 0.53 0.76 0.40 0.47

28 28 18 30 18 18 18 27 18 18 18

23 24 18 26 18 18 18 24 18 18 18

23 26 28 23 23 26 21 22 24 21 20

21 23 22 21 20 24 19 20 21 19 18

Table IV Intra-observer performance in sagittal tomography of 60 temporomandibular joints

Overall agreement Observer

(%)

Kappa index

AlA BIB

90 80 85 78 83

0.79 0.60 0.69 0.58 0.67

C/C DID Mean

Discussion Corrected sagittal tomography has been established in previous studies to be the method of choice for examination of the TMJI,3-5,13. Frontal tomography has been found to provide only minor additional information on degenerative TMJ diseases, 14 and was therefore omitted. As one of the aims was to study observer performance, high-quality tomograms were used in order to minimize the effect of other variables, such as density, contrast and spatial resolution which, in clinical practice, might influence the diagnostic accuracy. The information on prior probability was given in order to standardize the observers' decision criterion. We consider cyst, erosion, an osteophyte and sclerosis to be the significant features in establishing a diagnosis of TMJ disease. Akerman et alY have shown

Sagittal tomography of the TMJ: W. Cholitgul et al. that, microscopically, there is a loss of articular soft tissue in areas that correspond radiologically to a cyst or erosion, whereas in those with a concavity, flattening or an osteophyte this layer is still intact. With regard to osteophytes, the situation may be similar to other joints such as the knee; small osteophytes are often found without any apparent deterioration 16 whereas more extensive lesions are agreed to be indicative of degenerative changes!". This point, that it is not only the osteophyte per se but also its size, has not been taken into account in previous studies of the TMJ, but we considered it crucial in selecting our representative images. The diagnostic significance of sclerosis is more problematical. Lindvall et al. 6 regarded it as similar in value to an erosion or an osteophyte, but this was confirmed by Akerman et al. 15 to be valid only for the condyle but not the temporal component. The sensitivity and specificity were high, perhaps partly explained by the selection of only high-quality tomograms. Another factor was the use of reference tomograms. When these were used in a previous study by Rohlin et al. 8 , they found somewhat lower values for sensitivity and specificity. However, the present experiment was modified in one respect: the assessment of the reference tomograms was discussed among the observers who also knew that they had been verified microscopically. This difference might help explain the higher accuracy, as access to cognitive feedback has been shown to be highly effective for teaching diagnostic accuracy'", One of the observers had a lower sensitivity but a higher specificity than the others, demonstrating that these indices are interrelated and affected by a particular radiologist's tendency to either over- or under-read. Even though the educational background and experience in TMJ radiology of the observers were quite different, the diagnostic accuracy and the interobserver agreement were high. On the basis of the Kappa index, the interobserver a~reement was fair to substantial for any two observers 1 • This quite good result is probably due to the use of reference films, consistent with other radiographic studies on rheumatoid arthritis20-22 • Calibration of the observers by discussing the reference radiographs collectively appears to have been effective. In consequence, as others'"- 23, 24 have noted, there was little difference between intra- and inter-observer agreement. The presence of an osteophyte was reported accurately, due perhaps to their site on the condyle which is better visualized radiographically than the temporal component. The size of those selected for this study certainly increased the diagnostic accuracy. The interpretation of smaller lesions, on the other hand, might well cause problems, particularly in the central and medial thirds, where the insertion of the lateral pterygoid muscle on the anterior surface of the condylar head and neck can cause a bony beak, the socalled 'traction spur'. In the lateral third of the condyle the anterior surface should normally be rounded. Cyst or an erosion was a true positive finding in about twothirds of cases, whereas sclerosis was most frequently a false positive. This outcome may be explained by the difficulty in differentiating a rather wider layer of compact bone than usual from genuine sclerosis, even though a definition was agreed. Similar problems have been experienced with the histological evaluation:

sclerosis localized to the articular eminence is seldom associated with macroscopic changes'", For these reasons, in particular the high rate of false diagnoses, we suggest that the use of sclerosis as a sign of osteoarthrosis in the TMJ should be re-evaluated.

References 1. Blair GS, Chalmers 1M. Radiology of the temporomandibular joint. A comparison of circular tomography with orthopantomography and lateral tanscranio-oblique radiography. J Dent 1972; 1: 69-76. 2. Omnell KA, Petersson A. Radiography of the temporomandibular joint utilizing oblique lateral transcranial projections. Comparison of information obtained with standardized technique and individualized technique. Odontol Rev 1976; 27: 7792. 3. Petersson A. Radiography of the temporomandibular joint. A comparison of information obtained from different radiographic techniques [Thesis] Lund: University of Lund, 1976. 4. Stanson AW, Baker Jr HL. Routine tomography of the temporomandibular joint. Radiol Clin North Am 1976; 14: 10527. 5. Mongini F. The importance of radiography in the diagnosis of TMJ dysfunctions. A comparative evaluation of transcranial radiographs and serial tomography. J Prosthet Dent 1981; 45: 186-98. 6. Lindvall AM, Helkimo E, Hollender L, Carlsson GE. Radiographic examination of the temporomandibular joint. A comparison between radiographic findings and gross and microscopic morphologic observations. Dentomaxillofac Radiol 1976; 5: 2432. 7. Bean LR, Omnell KA, Oberg T. Comparison between radiologic observations and macroscopic tissue changes in temporomandibular joints. Dentomaxillofac Radio11977; 6: 90-106. 8. Rohlin M, Akerman S, Kopp S. Tomography as an aid to detect macroscopic changes of the temporomandibular joint. An autopsy study of the aged. Acta Odontol Scand 1986; 44: 131-40. 9. Petersson A, Rohlin M. Rheumatoid arthritis of the temporomandibular joint. Evaluation of three different radiographic techniques by assessment of observer performance. Dentomaxillofac Radio11988; 17: 115-20. 10. Romeis B. Mikroskopische technik, 16. Munich: Aufl. R Oldenburg Verlag, 1968: 164 pp. 11. Weinstein MC, Fineberg HV. Clinical decision analysis Philadelphia: Saunders, 1980. 12. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Measurement 1960; 20: 37-46. 13. Omnell KA. Radiology of the TMJ. In: Irby WB, ed. Current advances in oral surgery, vol III. St Louis: C.V. Mosby, 1980: 196-226. 14. Tanimoto K, Petersson A, Rohlin M, Hansson LG, Johansson CC. Comparison of computed with conventional tomography in the evaluation of temporomandibular joint disease: a study of autopsy specimens. Dentomaxillofac Radio11989; 18: 151-5. 15. Akerman S, Kopp S, Rohlin M. Macroscopic and microscopic appearance of radiologic findings in temporomandibular joints from elderly individuals. An autopsy study. Int J Oral Maxillofac Surg 1988; 17: 58-63. 16. Danielsson L, Hernborg J. Clinical and roentgenologic study of knee joints with osteophytes. Clin Orthop 1970; 69: 302-12. 17. Sokoloff L. The pathology of osteoarthrosis and the role of ageing. In Nuki G, ed. The aetiopathogenesis of osteoarthrosis. Kent: Pitman Medical, 1980: 1-15. 18. Wigton RS, Patil KD, Hoellerich VL. The effect of feedback in learning clinical diagnosis. J Med Educ 1986; 61: 816-22. 19. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159-74. 20. Rohlin M, Petersson A. Rheumatoid arthritis of the temporomandibular joint. Radiologic evaluation based on standard reference films. Oral Surg Oral Med Oral Patho11989; 67: 594-9.

Dentomaxillofac. Radiol., 1990, Vol. 19, February 5

Sagittal tomography of the TMJ: W. Cholitgul et al. 21. Kellgren JH, Lawrence JS. Radiological assessment of rheumatoid arthritis. Ann Rheum Dis 1957; 16: 485-93. 22. Larsen A, Dale K, Eek K. Radiographic evaluation of rheumatoid arthritis and related conditions by standard reference films. Acta Radiol (Diagn) 1977; 18: 481-91. 23. Blair GS, Chalmers 1M, Leggat TG, Buchanan WW. Circular tomography of the temporomandibular joint. A critical evaluation of the accuracy and reproducibility of the technique. Oral Surg Oral Med Oral Patho11973; 35: 416-27.

6 Dentomaxillofac. Radiol., 1990, Vol. 19, February

24. Kopp S, Rockier B. Variation in interpretation of radiographs of temporomandibular and hand joints. Dentomaxillofac Radiol 1978; 7: 95-102.

Address: Dr Madeleine Rohlin, Department of Oral Radiology, School of Dentistry, Carl Gustafs vag 34, S-214 21 Malmo, Sweden.