Comparisons of MRI Findings by Univariate Analysis
There was no significant difference in sex or age between the adhesive capsulitis group and the control group.
Results of comparison of MRI findings between the two groups are summarized in Table 1.
All qualitative and quantitative MRI findings were significantly different between adhesive capsulitis and control groups.
The mean anterior capsular thickness in the adhesive capsulitis group was higher than that of the control group (3.99 ± 1.64 mm vs.
1.66 ± 0.79 mm,
p < 0.001).
Anterior capsular edema was also more frequently observed in the adhesive capsulitis group (25 vs.
4,
p < 0.001).
There was no significant relationship between external rotation values measured in MR image and other MRI findings including anterior capsule thickness in the adhesive capsulitis group or the control group.
Multivariable Analysis and Diagnostic Performance
Among quantitative MR findings,
humeral and glenoid capsular thickness was excluded from multivariate analysis because the maximal axillary capsular thickness was more clinically used.
It is also correlated with humeral and glenoid capsular thickness.
Multivariate logistic regression analysis showed that anterior capsular thickness and maximal axillary capsular thickness were useful variables that could differentiate adhesive capsulitis from control group with odd ratios of 7.97 and 17.75,
respectively (p < 0.05).
Multivariate logistic regression analysis with anterior capsular edema,
axillary capsular edema,
and edema at the subcoracoid fat triangle in qualitative findings showed that only anterior capsular edema was significant,
with odd ratio of 12.41 (p = 0.01) (Table 1).
In ROC analysis,
anterior capsular thickness showed higher diagnostic performance than maximal axillary capsular thickness to diagnose adhesive capsulitis,
with AUC of 0.897 and 0.868,
respectively.
However,
anterior capsular thickness and maximal axillary capsular thickness in ROC comparison were not significantly different (95% confidence interval: 0.787-0.970 vs.
0.682-0.914,
respectively,
p = 0.28).
Results of cut-off value and area under the ROC curve are shown in Table 2 and Figure 9.
The cut-off value of anterior capsular thickness at 3.5 mm was clinically applicable,
affording excellent diagnostic accuracy,
with a sensitivity of 65.5% and a specificity of 100% (Fig.
10,
11).
The cut-off value of maximal axillary capsular thickness at 4 mm was also clinically applicable,
affording excellent diagnostic accuracy,
with a sensitivity of 58.6% and a specificity of 100%.
In five patient (17.2%) of adhesive capsulitis group,
anterior capsular thickness values were more than 3.5 mm.
However,
they did not meet the diagnostic criteria of maximal axillary capsular thickness (> 4 mm) and coracohumoral ligament thickness (> 3 mm).
The other qualitative parameters except anterior capsular edema were also negative in 2 (6.9%) out of these 5 patients (Fig.
12,
13,
14,
15).
Interobserver Agreement
Results of interobserver agreement are summarized in Table 3.
Good agreement was found between anterior capsular edema (κ = 0.77),
edema in humeral and glenoid capsules (both κ = 0.75),
humeral capsule thickness in axillary recess (ICC = 0.69),
and anterior capsule thickness (ICC = 0.66).
Moderate agreement was found for coracohumeral ligament thickness (ICC = 0.58),
glenoid capsule thickness in axillary recess (ICC = 0.47),
edema at the subcoracoid fat triangle (κ = 0.45),
and obliteration of the subcoracoid fat triangle (κ = 0.43).