*2.1. Study Population*

This retrospective observational study was approved by the institutional ethics committee, with a waiver of patient informed consent (CER-VD protocol number 505-15). We considered the following two patient groups, who did not need to be matched since they were analyzed separately according to the study design and primary objective (i.e., association between 3D acromion shape and glenoid retroversion).

The normal group included trauma patients aged 18 to 40, who had undergone a whole-body CT scan covering at least one of the two scapulae in full. Exclusion criteria were any radiological sign or history in medical records of disorders of the shoulder bones and joints (OA, fracture, glenoid dysplasia, or prior surgery of the upper limb), CT signs of immature skeleton (absence of fusion between any of the scapular ossification centers [16]) or CT artifacts (motion or metal). From our institutional picture archiving and communication system, an attending musculoskeletal radiologist retrospectively reviewed 221 consecutive whole-body CT scans performed over a 6-month period, and from these 112 patients met the inclusion criteria. The main characteristics of the normal subjects (79 males and 33 females) were mean age, 28.4 years (range, 18–40); mean height, 174.4 cm (range, 150–210 cm); mean weight, 75.5 kg (range, 50–120 kg); mean body mass index, 24.7 kg/m<sup>2</sup> (range, 18.6–38.1 kg/m2).

The pathological group consisted of patients with glenohumeral OA who had undergone a shoulder CT scan covering the entire scapula in their preoperative planning prior to shoulder arthroplasty. Patients with any traumatic injury to the shoulder girdle, malunion or nonunion, necrosis of the humeral head, or rheumatoid arthritis were excluded. Of the 334 consecutive patients eligible from 2002 to 2016, 125 with primary glenohumeral OA met the inclusion criteria. The main characteristics of the OA patients were mean age, 71.4 (range, 46–88 years); 37 males, 88 females; mean height, 165.7 cm (range, 141–186 cm); mean weight, 78.4 kg (range, 42–129 kg); mean body mass index, 28.5 kg/m<sup>2</sup> (range, 17.7–43.6 kg/m2). According to the updated Walch classification [17], the distribution of glenoid types was: A1, n = 26; A2, n = 23; B1, n = 26; B2, n = 37; B3, n = 8; C, n = 5.

## *2.2. CT Protocols*

All CT scans were performed on multidetector-row CT systems (8–256 detector rows) from the same manufacturer (GE Healthcare), with standardized data acquisition and image reconstruction settings. For normal subjects, scapular CT images were reconstructed as follows: section thickness, 1.3 mm; section interval, 0.7–1.3 mm; kernel, sharp (bone or bone plus); pixel size, 0.4–1.0 mm. For OA patients, shoulder CT images were reconstructed as follows: section thickness, 0.6–1.3 mm; section interval, 0.3–1.0 mm; kernel, sharp (bone or bone plus; GE Healthcare); pixel size, 0.3–0.6 mm.

#### *2.3. Scapular Coordinate System*

All CT scans were analyzed in 3D using a reliable semi-automated method providing a scapular coordinate system described in detail elsewhere [18,19]. Briefly, the medio-lateral (*z*) axis was along the scapular axis, defined by the line fitting five points placed along the supraspinatus fossa projected in the scapular plane. The scapular (i.e., ~"coronal") plane was defined by three landmarks: the trigonum spinae (TS), the angulus inferior (AI), and the most medial of the five points defining the medio-lateral axis (Figure 1; additional illustrations on the coordinate system can be found in [19]). The postero-anterior (*x*) axis (i.e., ~"sagittal" plane) was then defined as being perpendicular to the scapular plane and medio-lateral axis. The infero-superior (*y*) axis (i.e., ~"axial" plane) was perpendicular to the other two axes. The origin of the coordinate system corresponded to the spinoglenoid notch projected on the medio-lateral scapular axis.

**Figure 1.** Anatomical description of the scapular coordinate system (OXYZ), acromion landmarks (AA, AC), trigonum spinae (TS), angulus inferior (AI), posterior extension of the acromion (AAx), acromion posterior angle (APA), acromion tilt angle (ATA), acromion length angle (ALA), and glenoid retroversion angle (GRA). The three axes (*<sup>x</sup>*, *y*, and *z*) correspond to postero-anterior, infero-superior, and medio-lateral, respectively.

## *2.4. Acromion Landmarks*

Two specific acromion landmarks were placed manually on its 3D surface using the same software (Amira; Thermo Fisher Scientific) and method as above [19]: the acromion angle (AA) and the most anterior point of the acromioclavicular (AC) joint (Figure 1). These two landmarks were characterized by their three coordinates in the scapular coordinate system (AAx, AAy, AAz; and ACx, ACy, ACz). Because of the expected variability in scapular size among patients, these coordinates (distances) were normalized by the scapular height, defined by the infero-superior distance between AI (AIy) and the origin of the scapular coordinate system.
