183 / 460
S. Lippacher, H. Reichel, M. Nelitz
182
2b) Interobserver agreement, two-
grade analysis (comparison of all
possible pairs of raters)
Two-grade analysis of lateral X-rays showed at
the first reading an agreement between 60%
and 76%and at the second reading an agreement
between 52% and 76%. The interobserver
agreement of axial MR images was at the first
reading 86% to 96% and at the second reading
62% to 86%.
3a) Agreement of the radiographs
and MRI, four-grade analysis
Using the four-grade classification according
to D. Dejour the agreement of radiographs and
MRI at the first reading was 36 to 56%. At the
second reading an agreement of radiographs
and MRI between 28 and 54% was achieved.
3b) Agreement of the radiographs
and MRI, two-grade analysis
For the two-grade classification the agreement
of radiographs and MRI at the first reading
ranged from 66 to 82%. At the second reading
an agreement of radiographs and MRI of 64 to
74% was measured.
Because of unsatisfactory results in the
agreement of radiographs and MRI, an
additional regression analysis was performed:
Two mixed effects regression models were
fitted to investigate the influence of radiographs
and MRI.
In the first regression model, the influence of
the method (radiographs vs. MRI) on the
D. Dejour’s classification was investigated.
The result shows that the MRI has a higher
odds for a higher level in D. Dejour’s
classification (odds ratio 1.29; 95% CI: 1.01;
1.66, p=0.04). In the second regression model,
the influence of the method (radiographs vs.
MRI) on the two-grade classification (low-
grade and high-grade trochlear dysplasia) was
investigated. The result of the second model
shows that the MRI has a higher odds for high-
grade trochlear dysplasia (odds ratio 2.24; 95%
CI: 1.61; 3.13, p<0.01).
Discussion
The results of the current study show that inter-
and intraobserver agreement on D. Dejour’s
four grade classification of trochlear dysplasia
on the lateral radiograph and the MRI is
insufficient. In contrast the two-grade classifi
cation showed good to excellent agreement.
There are different possible reasons for these
unsatisfactory results. The surface geometry
of the articular cartilage cannot easily be
divided in four groups as there is a wide
variation of the anatomy of the trochlear
geometry. For example we found it difficult to
distinguish between type B and C dysplasia.
One of the rater indicated a flat, but slightly
descending trochlea as type B, whereas the
other raters interpreted the slightly descending
trochlea as hypoplastic and therefore classified
it as type C. Type C dysplasia is difficult to see
on lateral radiographs and often needs to be
defined on coronal cuts on MRI-scans, as in
different cases a small vertical join or cliff
pattern on MRI-scans may lead one rater to
classify it as type C and the other rater as type
D (fig. 3, 4).
In clinical practice it can be furthermore
difficult to achieve a true lateral radiograph
with superimposed dorsal condyles. But true
lateral radiographs are mandatory to confirm
the diagnosis of trochlear dysplasia. Koëter et
al. verified that minimal rotation aberrations
cause radiographs misdiagnosis of trochlear
dysplasia [15].
The results of the study confirm the conclusions
of Remy
et al.
[20] saying that the crossing
sign and the supratrochlear bump are the most
reproducible signs whereas the double contour
only showed agreements of 51%.
Because the lateral radiograph was inconsistent
in differentiating D. Dejour’s four grades of
trochlear dysplasia, the ability to distinguish