182 / 460
Intra- and Interobserver agreement of Dejour´s classification of trochlear dysplasia
181
In method 2 the axial MRI of the proximal
trochlea was used. Selected for this study was
the most proximal image that included the
entire width of the trochlea [9] (fig. 1).
The fifty lateral radiographs and the fifty MRI
images were read by four different observers,
two senior and two junior orthopaedic surgeons,
who were blinded to the clinical history. The
studies were read twice within a four week
time span. For the second reading the studies
were randomly arranged in a completely
different order.
The results were first analyzed for interobserver
agreement on the four classification types
between the four readers. To see if there was
better agreement if the dysplasia was divided
into two groups instead of four, the interobserver
agreement was then analysed when type A was
compared to type B, C and D combined.
Statistical analysis
Contingency tables were used descriptively for
intra- and interobserver agreement as well as
for agreement of lateral radiographs and MRI-
scans. For assessment of agreement, the
proportion of observed agreement (po) was
calculated incl. 95% confidence interval (CI).
Because of unsatisfactory results in the
agreement of radiographs and MRI, an
additional analysis was performed: Two mixed
effects regression models [16] were fitted to
investigate the influence of radiographs and
MRI on the D. Dejour’s four grades of
radiological criteria of trochlear dysplasia
(ordinal outcome) and on the two-graded
distinguishing between low-grade and high-
grade trochlear dysplasia (binary outcome). In
both models the method (i.e. radiographs/MRI)
was included as fixed effect and the rater was
included as random effect.
Results
The current study assesses how four blinded
independent readers agree on typing trochlear
dysplasia first using just the lateral view and
then using just the MRI image. This was done
twice with the studies being randomly shuffled
for the second reading.
In the following, important results are
summarized in terms of proportions of observed
agreement (po).
1a)
Intra-observer agreement,
four-grade analysis
When observers classified trochlear dysplasia
into D. Dejour’s four grades the intra- observer
agreement of the lateral radiographs between
the first and the second reading was 30% to
78%.Intra-observeragreementforclassification
into four grades for the two readings of the
axial MR images was 32% to 74%.
1b) Intra-observer agreement,
two-grade analysis
When grade A was compared to type B, C and
D combined to form low-grade and high-grade
classifications the intra-observer agreement on
the lateral radiograph between the first and the
second reading was 56% to 88% and 70% to
90% for the axial MR images.
2a) Interobserver agreement, four-
grade analysis (comparison of all
possible pairs of raters)
Four-grade analysis of interobserver agreement
of lateral X-rays showed at the first reading an
agreement ranged from 36% to 52% and at the
second reading an agreement ranged between
24% and 58%. Interobserver agreement of
axial MR images was at the first reading 42%
to 68% and at the second reading 28% to 40%.