Interobserver reliability was substantial or almost perfect
Interobserver kappa values were 0.88 to 0.92 and 0.62 to 0.74 for classification of the intra-articular EBG and extra-articular EBG in the lateral view. The kappa values were 0.79 to 0.93 for classification of the EBG in the superoinferior view into types A, B and C and 0.80 to 0.93 for classification of the shape of the articular surface of the humeral head adjacent to the EBG. These data indicated substantial or almost perfect reliability across the 3 observers.
Measurement and classification distribution of the EBG in cadaveric specimens
The length, width, depth, medial wall angle and total open angle of the EBG in cadaveric specimens are presented in Table 1. The classification distributions of the EBG are presented in Table 2. Lesions of the LBT were found in 2 specimens: one type Ⅰ lesion (Figure 7A) and one type II lesion (Figure 7B) based on the Curtis-Snyder classification[16], and one small tear (Figure 7A) and 1 medium tear (Figure 7B) based on the Post classification[17]. There were 2 subluxations among all 34 type B LBTs, and no dislocations of the LBT were found (Figure 7B, 7C, 7D). Consequently, only 2 lesions of the BPC, one type Ⅰ (Figure 8A) and one type II lesion (Figure 8B, 8C), were detected based on the Habermeyer classification[21, 22]. Both injured LBTs were subluxated and were type B and type C; meanwhile, both injured LBTs were type Ⅰ and type iii EBGs (Figure 7A, 7B). No mesotendon of the LBT was found within the EBG.
Distribution of EBG classifications in dry humeri
The distribution of EBG classifications in dry humeri are presented in Table 3 and Figure 9. For EBG classifications in the superoinferior view, most humeri were type A with a "U" shape (81/113, 73.45%), while few were type C with a “fishhook” shape (10/113, 8.85%) (Figure 5). Meanwhile, for classifications in the lateral view, most humeri were type ii with a “sphecidae” shape (57/113, 50.44%), while few were type iii with a “flare” shape (14/113, 12.39%) (Figure 4). Regarding the articular surface of the humeral head adjacent to the EBG, most humeri were type b with wavelike lines (71/113, 62.83%), while few were of type a with smooth arcs (42/113, 37.17%) (Figure 6).
Measurement of the EBG and its relationship with classification in the dry humeri
The width, depth, medial wall angle and total open angle of the EBG in the dry humeri are presented in Table 4. The total open angle was the largest (F=63.52; P<0.001) and the depth and medial angle were the smallest (F=79.59 and 36.37, respectively; both P<0.001) in type B, and type C had the smallest width (F=5.34, P=0.006). Additionally, the depth and medial angle were largest (F=16.89 and 5.85; P<0.001 and P=0.004, respectively) and the total open angle was the smallest in type ii (F=10.82, P<0.001). The width, depth, medial wall angle and total open angle of the EBG in dry the humeri were not significantly different between the 2 types classified based on the articular surface adjacent to the EBG (t=0.53, 0.68, 0.87, and 0.87; all P≥0.39).
Distribution of EBG classifications based on 3-D CT
The distributions of EBG classifications based on 3-D CT are presented in Table 5 and Figure 10. Nearly half of the classifications were type ii with a “sphecidae” shaped (134/278, 48.20%) EBG, while few were type iii with a “flare” shape (24/278, 8.63%) (Figure 4). Meanwhile, most were type A with a "U" shape (176/278, 63.31%), while few were type C with a “fishhook” shape (29/278, 10.43%) (Figure 5). Regarding the articular surface of the humeral head adjacent to the EBG, most classifications were type b with wavelike lines (166/278, 59.71%), while few were type a with smooth arcs (112/278, 40.29%) (Figure 6).
Measurement of the EBG and the distribution based on the classification of images
Data on the width, depth, and angles in the images are presented in Table 6. The width and total open angle were the largest (F=280.75 and 538.56, respectively; both P<0.001) and the depth and medial angle were the smallest (F=18.63 and 85.25, respectively; both P<0.001) in type B. Additionally, the depth and medial angle were the smallest (F=30.10 and 18.74, respectively; both P<0.001) and the width and total open angle were the largest (F=14.93 and 27.14; both P<0.001) in type iii. The width, depth, medial wall angle and total open angle of the EBG in the images were not significantly different among the types based on classification of the articular surface adjacent to the EBG (t=0.14, 0.44, 0.68, and 0.63, respectively; all P≥0.50).
Lesions of the LBT and BPC on MRI and their correlations with classifications
The distributions of the LBT and BPC lesions on MRI are presented in Table 6 and Figure 10. Most LBT locations within the EBG were normal (261/278, 93.88%) (Figure 7E), with 13 subluxations (Figure 7F) and 4 dislocations (Figure 7G). Interestingly, all of the subluxations and dislocations occurred in type B, and most lesions of the LBT and BPC occurred in type B (61/120, 50.83%; 65/117, 55.56%, respectively) (χ2=16.55 and 26.05; both P<0.001). Meanwhile, most lesions were type Ⅰ according to the Curtis-Snyder classification (90/121, 74.38%) and type II according to the Post classification (61/121) (Table 5). Most BPC lesions were type II (78/117, 66.67%), while few were type III (7/117, 5.98%) (χ2=110.80; P<0.001) (Figure 8). Notably, the LBT was injured in 83.56% of type B and 100% of type C EBGs (χ2=135.69, P<0.001), which warrants further investigation. In addition, the LBT was injured in 95.83% of type iii EBGs (χ2=31.89, P<0.001). Similarly, the BPC was injured in 89.04% of type B and 93.10% of type C EGBs (χ2=153.14, P<0.001), while the BPC was injured in 95.8% of type iii EBGs (χ2=33.79, P<0.001).