Minimal but potentially clinically relevant anteroinferior position of the humeral head following traumatic anterior shoulder dislocations: A 3D‐CT analysis

In unstable shoulders, excessive anteroinferior position of the humeral head relative to the glenoid can lead to a dislocation. Measuring humeral head position could therefore be valuable in quantifying shoulder laxity. The aim of this study was to measure (1) position of the humeral head relative to the glenoid and (2) joint space thickness during passive motion in unstable shoulders caused by traumatic anterior dislocations and in contralateral uninjured shoulders. A prospective cross‐sectional CT‐study was performed in patients with unilateral anterior shoulder instability. Patients underwent CT scanning of both injured and uninjured side in supine position (0° abduction and 0° external rotation) and in 60°, 90°, and 120° of abduction with 90° of external rotation without an external load. Subsequently, 3D virtual models were created of the humerus and the scapula to create a glenoid coordinate system to identify poster‐anterior, inferior‐superior, and lateral‐medial position of the humeral head relative to the glenoid. Joint space thickness was defined as the average distance between the subchondral bone surfaces of the humeral head and glenoid. Fifteen consecutive patients were included. In supine position, the humeral head was positioned more anteriorly (p = 0.004), inferiorly (p = 0.019), and laterally (p = 0.021) in the injured compared to the uninjured shoulder. No differences were observed in any of the other positions. A joint‐space thickness map, showing the bone‐to‐bone distances, identified the Hill−Sachs lesion footprint on the glenoid surface in external rotation and abduction, but no differences on average joint space thickness were observed in any position.

Proper centering of the humeral head in relation to the glenoid is dependent on the bony structures of the glenohumeral joint, muscle strength and coordination, tendon laxity, capsule, and ligaments.
When a force is applied to the humeral head, it can result in humeral translation, and this may be a suitable parameter to quantify laxity of the shoulder.A shoulder dislocation can occur when there is too much translation due to failure of compensation mechanisms, resulting in impaired quality of life, kinesophobia, and shoulder pain at least up to 12 months after the injury. 1 Moreover, recurrence of (multiple) dislocation(s) occurs in up to 60% of patients younger than 40 years following nonoperative treatment. 2If patients are expected to have a high risk of recurrence, operative treatment can be performed to reduce recurrence risk.To facilitate weighing recurrence and complication risks and improve patient selection, multiple methods were proposed to estimate recurrence risk.4][5] However, even the glenoid track concept shows differences in effect size, caused by patient selection and not being able to include muscle strength and coordination, tendon laxity, and damage to soft-tissue structures in the formula. 5n Eisenhart-Rothe et al. demonstrated that the humeral head is positioned more anteriorly and inferiorly relative to the glenoid in passive 90°of abduction with/without 90°of external rotation following unilateral traumatic anterior shoulder dislocations compared to the contralateral uninjured side through 3D-MRI analysis. 6ey also demonstrated that muscle activity in 90°of abduction with 90°of external rotation slightly recentered the humeral head.

Matsumura et al. evaluated humeral translation during active external
rotation in 90°of abduction and found decreased posterior, inferior and medial translation following unilateral traumatic anterior shoulder dislocations compared to the contralateral side through 4D-CT analysis.Without external load, the differences in both studies include only a few millimeters, which may be difficult to correlate to recurrence risk due to measurement errors.To accurately measure these small differences and prevent measurement errors, a standardized measurement that also includes reliability and assessment of the measurement error has to be performed.As alternative for position measurements, the centering of the humeral head and cartilage thickness can be evaluated by determining the distance between the subchondral bone surfaces of the humeral head and the glenoid, or joint space thickness, which is a technique that has already been used in the wrist. 7If the humeral head moves posterior-anteriorly or inferior-superiorly, the distance between the bony structures is expected to decrease due to the concave shape of the glenoid.This technique can potentially visualize in which direction the humeral head is moving.
If the position of the humeral head can be measured reliably, then this may serve as a quantitative measure of shoulder laxity.A standardized evaluation that takes both the position of the humeral head and joint space thickness into account can provide useful information of the centering of the humeral head relative to the glenoid in passive motion in shoulder instability patients.This can help us to (1) develop a more integrated approach for quantification of shoulder instability based on multiple factors and (2) determine to what extent it can be managed with surgical treatment in the future.
Therefore, the aim of this study was to measure (1) position of the humeral head relative to the glenoid and (2) joint space thickness during passive motion in unstable shoulders caused by traumatic anterior dislocations and in contralateral uninjured shoulders.It was hypothesized that unstable shoulders would be positioned more anteriorly in external rotation and abduction, as was observed by Von Eisenhart-Rothe et al. and Matsumura et al. 6,8 In addition, it was hypothesized that a decreased joint space thickness is observed in these positions caused by anterior translation and the concave shape of the glenoid. 9

| METHODS
This prospective cross-sectional study with level of evidence II was approved by the Amsterdam Medical Ethics Committee (METC AMC) under registration number NL66670.018.18 and the local medical ethics committee of the OLVG.This manuscript was written according to the Strengthening The Reporting of Observational Studies in Epidemiology (STROBE) checklist for cross-sectional studies. 10

| Study participants
Patients between 18 and 50 years old that experienced a traumatic anterior shoulder dislocation, that were planned for arthroscopic Bankart repair and would receive a preoperative MRI, who experienced two or more complete shoulder dislocations in one shoulder requiring closed reduction by a physician and had no dislocations in the contralateral shoulder, and a proper understanding of the Dutch or English language were included in this study.Patients with a full-thickness rotator cuff tear, positive sulcus sign (>1 cm translation), joint hypermobility (Beighton score >2), severe osteoarthritis (Samilson-Prieto >2), or a (pathological) fracture other than a Hill−Sachs lesion or glenoid bone loss were excluded. 11Also patients without a Hill−Sachs lesion on MRI-to ascertain that patients experienced anterior shoulder dislocation as not all patients had a radiograph with dislocated shoulder available-and patients that received previous shoulder surgery were excluded.

| Image acquisition and 3D-segmentation
CT images were acquired using the following parameters for a regular-dose scan: 250 mAs, 120 kV, and slice thickness 1.0 mm.A regular-dose scan was performed of both shoulders with the patient in supine position (0°abduction and external rotation), with the arms alongside the body.Subsequently, the patient was placed on a plywood plate to facilitate positioning in the CT scanner by a flat surface to ensure standardization of the position of the arm during CT-scanning (Figure 1A).A holder (1B) was used to hold the wrist and position the arm on the plywood plate.While the patient was laying on the plywood plate, the apprehension and relocation tests were performed by an investigator trained in assessment for anterior shoulder instability (L.P. E. V.). 12,13Six low-dose CT-scans (60 mAs, 120 kV, and slice thickness 1.0 mm) were performed with the arm in 90°of external rotation and 60°, 90°, and 120°of abduction on the injured and uninjured side.The patient was placed in these positions by use of a goniometer to determine the angle with the plywood plate as reference.The total radiation dose received by the patient was estimated to be 2 mSv.The regulardose CT image was segmented using a custom-made software package to create 3D models of the humerus and scapula.This method of segmentation was validated previously. 14,15The 3D model of the regular-dose CT scan was aligned with the low-dose CT scans by using a process of registration to find the position and orientation of the humeral head with respect to the scapula. 7,15is method of a regular-dose CT scan for segmentation of the bones, followed by registration to the low-dose CT scans, allows for multiple scanning with reduced radiation exposure.[18][19] 2.3 | Definition of the glenoid coordinate system and of the humeral position Consensus on definition of a glenoid coordinate system has yet to be reached.Similar to Matsumura et al., the glenoid coordinate system was defined as a Cartesian coordinate system with three orthogonal axes. 8The following definitions were chosen as they were thought to be least influenced by morphological differences between the injured and uninjured glenoid: the origin was defined as the centroid of the selected articular surface of the glenoid.The Z-axis was defined perpendicular to a best-fitted plane of the selected articular surface of the glenoid, as this has shown to have a good reliability (Figure 2B). 6,8,20A line was drawn between the most superior and most inferior point of the glenoid surface and the X-axis was perpendicular to both the Z-axis and the temporary line.The Y-axis was perpendicular to both the Z-axis and X-axis.Translation along the X-axis was defined as posterior (negative) and anterior (positive), translation along the Y-axis was defined as inferior (negative) and superior (positive), and translation along the Z-axis was defined as medial (negative) and lateral (positive).The center of the humeral head was defined as the center of a best-fitted sphere of the articular surface of the subchondral bone, similar to previous studies (Figure 2A). 6,21,22lecting the articular surface was based on changes in the local curvature of the joint surface, of which a detailed explanation is added in Supporting Information 1. Sphere fitting was performed The patient was placed on a plywood plate (A) to facilitate positioning in the CT scanner by a flat surface and to ensure standardization of the measurements.The patient would lie supine on the plywood plate and was positioned with the wrist in a holder (B) by the lead investigator who measured the shoulder angle with a goniometer.The back of the plywood plate contained a mirrored image to facilitate standardized measurement on the contralateral side.automatically on the selected articular surface and was based on the iterative method described by Eberly. 23Humeral head position was defined as the center of the humeral sphere in the Cartesian coordinate system of the glenoid (Figure 2C).Humeral translation was defined as the displacement of the center of the humeral head in the glenoid coordinate system from supine position (0°external rotation and abduction) to the 60°, 90°, or 120°abduction position with the arm in 90°of external rotation.

| Joint space thickness map
Joint space thickness at each position was defined as the average distance between the subchondral bone surface of the humeral head and the subchondral bone surface of the glenoid by using the joint space distance map, as previously specified for the wrist. 7,24In short, for each point in the glenoid surface, a line in the direction of the surface normal (vector perpendicular to the bony surface) was determined.The distance between the glenoid point and the point where the line intersects the opposite surface was assigned to each glenoid point.Two pragmatic restrictions were applied to find the average point distance, including (1) a maximum distance between the humeral head and glenoid of 15 mm and (2) a maximum angular difference of 60°between the normal vectors of each glenoid point and the normal vector of the opposing point. 7,24To visualize the distances, a joint space thickness map was created by assigning colors to the distances.

| Measurement errors
To determine the segmentation and registration error of the custommade software for the humerus and scapula, a single whole cadaver was scanned in three different positions with the same CT-scanner and imaging specifications.The three positions included 90°external rotation and (1) 60°, (2) 90°, and (3) 120°of abduction.Next to a regular-dose scan (250 mAs) for segmentation, the cadaver was scanned five times in each position in with a low-dose (60 mAs) scan to evaluate the registration error.For each regular-dose scan, the process of creating 3D humeral and scapular polygons was repeated five times by the lead author (L.P. E. V.).The first segmented polygon for the humerus and the scapula served as a reference polygon.The segmentation error was defined as the distance between each point that formed the 3D polygon and the nearest point of the reference polygon.To determine the interobserver measurement error, three randomly chosen glenohumeral joints were used to repeat the entire measurement from segmentation to determining the position of the center of the humeral head in the glenoid coordinate system by the lead author (L.P. E. V) and an independent colleague from the department that had experience with the software, but was not further involved in the study.Both observers used the same settings within the software.This measurement was repeated at least 2 weeks later to determine the intraobserver measurement error.The observers were blinded to each other's measurements.An average measurement error was calculated by determining the difference between observers for each observation (interobserver) and between time points (intraobserver).
The center of the humeral head was defined as the center of a best-fitting sphere of the humeral articular surface (A).The Z-axis (lateral-medial) was defined as the line perpendicular to a best-fitting glenoid plane (B).The X-axis (posterior-anterior) was defined as perpendicular to the Z-axis and a line between the most superior and inferior points (indicated with yellow cross) of the glenoid (line not shown).The Y-axis (inferior-superior) was perpendicular to both the X-and Z-axes, completing the Cartesian glenoid coordinate system (C).3 | RESULTS

| Participants
Twenty-four consecutive patients were assessed for eligibility of which 15 patients with unilateral traumatic anterior shoulder dislocations matched the inclusion criteria, agreed to participate, and were included between October 15, 2020 and November 19, 2021.The most important reason for the nine patients not to participate was the travel distance for the additional CT scan.All participants in this study provided written informed consent and were included at the OLVG hospital in Amsterdam (Table 1).The included patients had an average age of 30.2 ± 4.7 years and 67% was male.The right side was injured in 80% of the cohort.The first dislocation was experienced at an average age of 22.5 ± 4.7 years and the patients experienced a median of 4 (IQR 2.5) dislocations before they received a CT-scan for this study.

| Humeral position and translation
The position of the center of the humeral head in supine position (0°e xternal rotation and abduction) was more anterior in the injured shoulder F I G U R E 3 Sagittal view of the glenoid (not the true scale of the glenoid, but just to illustrate the view) in the anatomical coordinate system, with a scatter plot of the position of the humeral head relative to the centroid of the glenoid coordinate system in millimeters for the posterioranterior direction (horizontal axes) and the inferior-superior direction (vertical axes) in supine position.The individual data points are plotted for the injured (orange) and uninjured (blue) shoulder.The data points of the injured and uninjured shoulder are connected through a dotted line for each patient.In addition, the 95% confidence interval in both directions was plotted, which is shown as ellipsoid for the injured (orange) and uninjured (blue) shoulder.ANT, anterior; INF, inferior; POS, posterior; SUP, superior.
(1.5 ± 1.5 mm) compared to the uninjured shoulder (0.2 ± 0.9 mm; p = 0.004; Figure 3).The position was more inferior (median −0.5 ± 1.0 vs. 0.1 ± 0.9 mm; p = 0.019) and lateral (25.6 ± 2.2 vs 25.0 ± 2.2 mm; p = 0.021) in the injured shoulder compared to the uninjured shoulder.There was no difference in position of the center of the humeral head in any of the directions between the injured shoulder and uninjured shoulder for the positions 60°, 90°, and 120°abduction (Figure 4).The injured shoulder demonstrated more posterior translation compared to the uninjured shoulder and this was statistically significant from supine to 90°(p = 0.012) and from supine to 120°(p = 0.019) of abduction (Figure 5).In addition, it demonstrated less inferior translation between supine position and 60°of abduction (p = 0.002).No statistically significant differences in translation could be observed in the other positions.When comparing patients with a positive 12 to patients with a negative apprehension test, 3 there was no difference in position of the center of the humeral head in the posterior-anterior (p = 0.967), inferiorsuperior (p = 0.462), and lateral-medial (p = 0.230) directions in 90°of abduction in the injured shoulder.

| Joint space thickness
No difference in mean joint space thickness could be observed between the humeral head and the scapula in supine position between the injured (5.08 ± 0.63 mm) and uninjured (5.38 ± 0.33 mm; p = 0.077) shoulders.In addition, no difference could be observed in the 60°(5.46± 0.58 vs. 5.24 ± 0.47 mm; p = 0.168), 90°(5.53± 0.64 vs. 5.41 ± 0.60 mm; p = 0.594), and 120°(5.50± 0.69 vs. 5.31 ± 0.49 mm; p = 0.358) abduction positions.The joint-space thickness map visualized the distances between the humeral head and the glenoid (Figure 6), which demonstrated irregularities in the color distribution that was caused by the Hill−Sachs lesion (Figure 7).When comparing the joint-space thickness maps between patients, it was observed that there were differences in location of the Hill−Sachs lesions footprint on the glenoid and the pattern in the different positions (Supporting Information 2).In addition, the projection of the humeral head on the glenoid demonstrated that, regardless of the same humeral head position while in supine position, in some patients the humeral head moved differently relative to the glenoid on the injured side.

| Measurement errors
The cadaver experiment demonstrated an average segmentation error of 0.15 ± 0.16 mm for the humerus and 0.15 ± 0.22 mm for the scapula.
An average registration error of 0.04 ± 0.02 mm was observed.An average interobserver measurement error of 0.10 ± 0.06 mm was observed for posterior-anterior, 0.09 ± 0.09 mm for inferior-superior Sagittal view of the glenoid (not the true scale of the glenoid, but just to illustrate the view) in the anatomical coordinate system, with a scatter plot of the position of the humeral head relative to the centroid of the glenoid coordinate system in millimeters for the posterioranterior direction (horizontal axes) and the inferior-superior direction (vertical axes) in all positions.The individual data points are plotted for the injured (orange) and uninjured (blue) shoulder of each patient.In addition, the 95% confidence interval in both directions was plotted, which is shown as ellipsoid for the injured (orange) and uninjured (blue) shoulder.ANT, anterior; INF, inferior; POS, posterior; SUP, superior.and 0.74 ± 0.42 mm for lateral-medial directions.An average intraobserver measurement error of 0.11 ± 0.05 mm (ICC: 0.95) was observed for posterior-anterior, 0.11 ± 0.06 mm (ICC: 0.99) for inferior-superior, and 0.42 ± 0.31 mm (ICC: 0.98) for lateral-medial directions.

| DISCUSSION
The goal of this cross-sectional study was to evaluate the centering of the humeral head in patients with traumatic anterior shoulder instability by comparing the humeral position relative to the glenoid and the joint space thickness between the injured and contralateral uninjured shoulder in passive motion.The most important findings were that (1) in supine position (0°of external rotation and abduction) the humeral head is positioned more anterioinferiorly and laterally in the injured shoulder compared to the uninjured shoulder, with the largest difference in the posteroanterior direction; (2) no differences were observed in the position of the humeral head in 60°, 90°, or 120°of abduction between the injured and uninjured shoulder; (3) the injured shoulder demonstrated more posterior and less inferior translation when moving to 60°, 90°, or 120°of abduction from supine position; (4) a difference in position between shoulders with a positive and negative apprehension test in 90°of external rotation and abduction could not be observed; (5) no difference in mean joint space thickness between the humeral head and scapula was observed in any of the positions; (6) the Hill−Sachs footprint can be visualized on the glenoid surface using a joint space thickness map; and (7) in some patients the orientation of the humeral head relative to the glenoid was different on the injured compared to the uninjured side.The hypotheses that the humeral head shows more anterior position and translation and decreased joint space thickness is not confirmed for passive external rotation and 60°, 90°, and 120°abduction.
Given the results of this study, the hypothesis that the humeral head is positioned more anteriorly in 90°of abduction and external rotation as found by Von Eisenhart-Rothe et al. and Matsumura et al.
is not supported. 6,8Von Eisenhart-Rothe projected the center of the humeral head on a plane that was fitted on the glenoid and defined their glenoid coordinate system differently, which might explain part of the differences that were observed. 6In their study with healthy volunteers, they demonstrated that in passive motion the average humeral head position was anterior to the origin by the same degree (approximately 2 mm) as the patients that suffered from a traumatic dislocation. 6,21This means that even though the values are different, the observed pattern may be the same and the difference may be explained by the definition of the coordinate system.There are no data of the 0°abduction and external rotation position which could be compared between both studies to test this hypothesis.whereas the current study did not find this difference. 8This might be explained by the dynamic aspect of their study, in which patients were asked to actively move from 0°to maximum external rotation.
Active movement and the difficulty of standardizing this measurement could have contributed to the differences observed between studies.Von Eisenhart-Rothe et al. already showed that there might be differences in humeral head position between passive and active motion. 6All patients in the study by Matsumura et al. reported a feeling of apprehension, which might contribute to changed kinematics during active movement to prevent pain or a feeling of apprehension and influence position measurements as well.An explanation for the result that the shoulder is positioned more anteriorly in supine position is that the CT-bench is slightly curved and applies a small anteriorly directed force on the shoulder.
Subsequently, the uninjured shoulder can prevent anterior translation, whereas the injured shoulder is not due to damage to the shoulder.It could be interesting to repeat the measurement, while applying more anteriorly directed force to the shoulder from behind to try and increase the difference.As expected, this study observed a joint space thickness that is slightly larger than the reported combined cartilage thickness of the glenoid and humerus. 25An explanation for finding no differences on average joint space thickness in supine position might be that the differences in position are too small to be identified through this analysis as it is overshadowed by using the average of all joint space thickness measurements.The joint-space thickness map showed that there are indeed differences created by the Hill−Sachs lesion in external rotation and abduction positions.The differences in humeral F I G U R E 6 Visualization of the joint space thickness between the humeral head and the glenoid subchondral bone for the injured and uninjured shoulder in all positions through a joint space thickness map in patient HT013.The humeral head is transparent and was projected on the glenoid.A red color indicates a small distance between the bony structures and a green color a larger distance.The injured shoulder demonstrates an area with irregular green color distribution that moves downwards as the shoulder is abducted (indicated by black arrows), which is caused by the Hill−Sachs lesion.
orientation that was observed in some patient could indicate dyskinesia.
An off-track lesion is currently the most consistent risk factor that is associated with recurrence in shoulder instability research. 4,5wever, the risk ratio varies between studies, which may be explained by observing only the bony features of the glenoid track and not the overlying cartilage or other fibrous tissue. 5The position of the humeral head relative to the glenoid in supine position might be a potential risk factor for redislocation, as it can be seen as a result of shoulder laxity or muscle activity patterns.Future research could therefore focus on the difference that is observed in humeral head position in supine (0°external rotation and abduction) position and determine if it is associated with a higher recurrence risk.If increased anterior position is associated with recurrence, it can be used as a risk factor.It would be interesting to see if patients that demonstrate a more anterior position in 3D could be treated by performing surgical treatment and if treatment is effective in these patients.As the observed difference was very small, future studies could determine if a different position of the shoulder or applying force to the shoulder from behind in the CT scan would lead to a larger difference in anterior position.For example, will it increase with the patient lying on the stomach or standing up or with a pad placed under the shoulder?

| Limitations
The results and conclusions should be considered by taking the limitations of the study into account: The directions (X, Y, Z) are based on a glenoid coordinate system for which a consensus has yet to be reached. 6,22,26Different glenoid coordinate systems are used in the literature to evaluate centering of The Hill−Sachs lesion faces the glenoid surface in the external rotation and abduction positions, demonstrating that the irregularities in the joint-space thickness map are caused by more distance between the bony structures where the Hill−Sachs lesion is facing the glenoid surface.
7][28] The glenoid plane has been demonstrated to have good reliability, but the position of the temporary line that was drawn between the most superior and inferior part of the glenoid may be vulnerable to changes due to large defects of the glenoid.This may influence the position measurements. 20,22It is unclear to what degree the direction of the glenoid coordinate system is influenced by the variation within a patient.
Differences of only a few millimeters were found and average measurement errors of 0.1-0.4mm were observed despite the standardized measurement.The inter/intraobserver measurement error exceeded the difference found in the lateral direction and should therefore be interpreted with caution.Small differences warrant reliable measurements, and this can be improved by minimizing manual interactions through improved automatic segmentation and 3D measurements.
Some patients were unable to reach the 120°position or the full 90°of external rotation caused by unbearable apprehension.
Furthermore, only 15 patients were included in this study and probably more patients need to be added to draw firm conclusions about these small differences.
The study only focused on the position of the humeral head relative to the glenoid and did not include severity of anterior glenoid bone loss or Hill−Sachs lesion, scapula position relative to the body or scapula morphology in the analyses.
In conclusion, compared to the contralateral shoulder, the humeral head demonstrated minimal anteroinferior position relative to the glenoid in supine unloaded position following traumatic anterior shoulder dislocations.This difference is potentially clinically relevant as it may serve as risk factor for redisloctation that is the result of shoulder laxity.No difference on average joint space thickness was observed in any position, but the joint-space thickness map of the affected joints identified the Hill−Sachs footprints on the glenoid surface in external rotation and abduction positions.

T A B L E 1 2 . 6 |
Abbreviations: A, ambidexterity; ABD, abduction; EXO, exorotation; F, female; L, left; M, male; R, right.a Complete dislocations.bThe patient experienced too much anxiety/pain to perform the scan in this position.
Matsumura et al. found the humeral head to be positioned more anteriorly in 90°of abduction and above 20°of external rotation, F I G U R E 5 Difference between the injured and uninjured shoulder expressed as translation of the humeral head with respect to supine (0°) position, for 60°of abduction (green), 90°of abduction (yellow), and 120°of abduction (gray).This was achieved by subtracting translation in the injured shoulder from the translation in the uninjured shoulder.(*) indicates statistical significance; (¥) Indicates in which direction the shoulder was moving.