Impact of Sagittal Rotation on Axial Glenoid Width Measurement in the Setting of Glenoid Bone Loss
Authors’ Disclosure Statement: The authors report no actual or potential conflict of interest in relation to this article.
Dr. Frank is Assistant Professor, Department of Orthopaedic Surgery, University of Colorado School of Medicine, Aurora, Colorado. Mr. Golijanin is a Medical Student, Geisel School of Medicine, Dartmouth Medical School, Hanover, New Hampshire. Dr. Vopat is Assistant Professor, Department of Orthopaedic Surgery, University of Kansas, Kansas City, Kansas. Dr. Gross is a Resident, DGMD Medical, Omaha, Nebraska. Dr. Chauhan is a Resident, Ninewells Hospital, Dundee, United Kingdom. Dr. Romeo is Chief of Orthopaedics, Rothman Institute New York, New York. Dr. Provencher is an Orthopaedic Surgeon, The Steadman Clinic, Vail, Colorado.
Address correspondence to: Rachel M. Frank MD, Department of Orthopaedic Surgery, University of Colorado School of Medicine, Aurora, CO 80045 (email, Rachel.Frank@ucdenver.edu).
Am J Orthop. 2018;47(6). Copyright Frontline Medical Communications Inc. 2018. All rights reserved.
Standard 2-dimensional (2-D) computed tomography (CT) scans of the shoulder are often aligned to the plane of the body as opposed to the plane of the scapula, which may challenge the ability to accurately measure glenoid width and glenoid bone loss (GBL). The purpose of this study is to determine the effect of sagittal rotation of the glenoid on axial anterior-posterior (AP) glenoid width measurements in the setting of anterior GBL.
Forty-three CT scans from consecutive patients with anterior GBL (minimum 10%) were reformatted utilizing open-source DICOM software (OsiriX MD). Patients were grouped according to extent of GBL: I, 10% to 14.9% (N = 12); II, 15% to 19.9% (N = 16); and III, >20% (N = 15). The uncorrected (UNCORR) and corrected (CORR) images were assessed in the axial plane at 5 standardized cuts and measured for AP glenoid width.
For groups I and III, UNCORR scans underestimated axial AP width (and thus overestimated anterior GBL) in cuts 1 and 2, while in cuts 3 to 5, the axial AP width was overestimated (GBL was underestimated). In Group II, axial AP width was underestimated (GBL was overestimated), while in cuts 2 to 5, the axial AP width was overestimated (GBL was underestimated). Overall, AP glenoid width was consistently underestimated in cut 1, the most caudal cut; while AP glenoid width was consistently overestimated in cuts 3 to 5, the more cephalad cuts.
UNCORR 2-D CT scans inaccurately estimated glenoid width and the degree of anterior GBL. This data suggests that corrected 2D CT scans or a 3-dimensional (3-D) reconstruction can help in accurately defining the anterior GBL in patients with shoulder instability.
- Standard 2-D CT scans of the shoulder are often aligned to the plane of the body as opposed to the plane of the scapula, which may challenge the ability to accurately measure glenoid width and GBL.
- Underestimations of GBL may lead to insufficient treatment of clinically meaningful GBL, thereby increasing the risk of instability recurrence; whereas overestimations of GBL may lead to unnecessary treatment, subjecting patients to increased surgical morbidity.
- AP glenoid width was consistently underestimated in uncorrected axial cut 1, the most caudal cut.
- AP glenoid width was consistently overestimated in uncorrected axial cuts 3 to 5, the more cephalad cuts.
- CORR 2-D CT scans or a 3-D reconstruction can help in accurately defining the anterior GBL in patients with shoulder instability.
The treatment of glenohumeral instability has substantially evolved over the past several decades. The understanding of glenoid bone loss (GBL), in particular, has advanced to such a level that we utilize the quantification of GBL for surgical decision-making. Unrecognized and/or untreated GBL is associated with recurrent instability, pain, and disability. Controversy exists, however, regarding the precise amount of anterior GBL that is significant enough to warrant surgical treatment. While historically, 25%1,2 of anterior GBL was thought to be the critical number required to warrant osseous augmentation, studies that are more recent have highlighted the need to perform osseous glenoid reconstruction with lesser degrees of GBL, particularly in the contact athlete.3-9 As small differences in the amount of GBL can change surgical decision-making from an all-soft tissue repair to an osseous reconstruction, it is paramount that we have accurate, valid, and reproducible methods for calculating GBL.
Historically, plain radiographs have been the mainstay for evaluating the glenohumeral joint, including Grashey and axillary views, allowing clinicians to evaluate the congruency of the glenohumeral joint and to assess bone loss on both the glenoid and humeral head.1,10 While large, acute fractures of the glenoid are fairly evident on radiographs, including the Grashey view,11 shoulders with chronic and/or attritional anterior GBL are more difficult to evaluate, and often do not provide the information necessary to guide surgical decision-making.
Computed tomography (CT) of the shoulder has become the most commonly utilized imaging modality in the evaluation of patients with shoulder instability associated with GBL. Standard 2-dimensional (2-D) CT scans of the shoulder are often aligned to the plane of the body as opposed to the plane of the scapula/glenoid, as standard protocols often fail to account for the anterior sagittal rotation of the scapula/glenoid, similar to the disadvantage of standard radiographs. While 3-dimensional (3-D) CT reconstructions eliminate the effect of gantry angles, and thus allow for an en face view of the glenoid, 3-D reconstructions are not always available, and cannot always be measured.12-14 Thus, improved methodology for utilizing standard 2D scans is warranted, as the ability to correctly align the axial CT scan to the axis of the glenoid may allow for more accurate GBL measurements, which will ultimately impact surgical decision-making. Recently, Gross and colleagues15 reported the effect of sagittal rotation of the glenoid on axial measurements of anterior-posterior (AP) glenoid width and glenoid version in normal glenoids, without bone loss, and found that the mean angle of correction needed to align the sagittal plane was 20.1° ± 1.2° of rotation. To the authors’ knowledge, this same methodology has not been applied to patients with clinically meaningful anterior GBL. Given that the average glenoid width in human shoulders is 24.4 mm ± 2.9 mm,16 1 mm of glenoid bone loss (GBL) corresponds to approximately 4% of the glenoid width, and thus even subtle differences in the interpretation of GBL may have substantial clinical implications. Therefore, the purpose of this study is to determine the effect of sagittal rotation of the glenoid on axial AP glenoid width measurements in the setting of clinically significant anterior GBL.
This study was approved by Massachusetts General Hospital Institutional Review Board. A retrospective review of consecutive patients with a diagnosis of anterior shoulder instability between 2009 and 2013 was conducted. Inclusion criteria comprised patients with a minimum of 10% anterior GBL, an available CT scan of the affected shoulder, and no history of prior ipsilateral surgeries. Exclusion criteria comprised evidence of degenerative changes to the glenoid and/or humeral head, as well as prior ipsilateral shoulder surgery. Sixty consecutive patients were originally identified as having anterior shoulder instability, and 17 were excluded based on the inclusion/exclusion criteria, leaving 43 patients (43 shoulders) available for inclusion. Shoulder CT scans from all 43 patients were reformatted utilizing open-source DICOM software (OsiriX MD, version 2.5.1 65-bit) multi-planar reconstruction (MPR).
All patients underwent a standard glenohumeral CT scan using a Siemens Sensation 64 Scanner (Siemens), a 64-detector scanner. Scans were acquired with 0.6 mm of collimation, 140 kV, and 300 mA-seconds. Slice thickness was set to 2 mm. All patient information was de-identified for analysis.
The uncorrected (UNCORR) scans were defined as the default orientation on the scanner. In the UNCORR scans, the axial, coronal, and sagittal views were oriented relative to the scanner gantry table, as opposed to the anatomy of the glenoid. The corrected (CORR) CT scans were aligned in all 3 planes relative to the glenoid face, and thus the cuts were perpendicular to the long axis of the glenoid.15 This resulted in sagittal cuts perpendicular to the 12-o’clock to 6-o’clock axis in the sagittal plane (Figure 1).
Image Analysis and Reformatting
In a de-identified fashion, all CT scans were imported and analyzed using open-source Digital Imaging and Communications in Medicine (DICOM) software (OsiriX MD, version 2.5.1 64-bit). By following a previously developed method, CT scans were reformatted using OsiriX MPR. The OsiriX software has an MPR function that allows simultaneous manipulation of 2-D CT scans in 3 orthogonal planes: axial, sagittal, and coronal. In the MPR mode, the alternation of 1 plane directly affects the orientation of the remaining 2 planes. Thus, by using an MPR, one can analyze the impact that a default CT scan performed relative to the gantry of the table, UNCORR, has on the axial images.
First, the en face view was obtained via a 2-step process: alignment of the axial plane to account for the scapular angle, followed by alignment of the coronal plane to adjust for the glenoid inclination.15 These 2 adjustments provided a true en face sagittal glenoid view. The final adjustment step was a sagittal en face rotation of the glenoid such that the superior and inferior glenoid tubercles were placed on the 12-o’clock to 6-o’clock axis (CORR scan). Previous studies have identified a central longitudinal axis that was used in this method to align the supraglenoid tubercle with the 12-o’clock to 6-o’clock axis on the glenoid face.15,17,18 The standard error of mean was 1.21°. This new CORR view resulted in axial cuts through the glenoid that were oriented perpendicular to the 12-o’clock to 6-o’clock axis. The UNCORR and CORR images were assessed in the axial plane at 5 standardized cuts and measured for AP glenoid width by 2 independent observers in a blinded, randomized fashion. When the measured AP width of the UNCORR scan was less than that measured on the CORR scan, the AP width of the glenoid was considered underestimated, and the degree of GBL was considered overestimated (Figure 2).
Scapular angle measurements were performed on the axial view as the angle between a line through the long axis of the body of the scapula, and a line parallel to the CT gantry table.15,19 Subsequently, the axial plane was aligned to the glenoid surface.
Coronal inclination measurements were performed on the sagittal view as the angle between a line tangential to the face of the glenoid and a line perpendicular to the CT gantry table. Positive values represented superior inclination, while negative values represented inferior glenoid inclination.15
Sagittal rotation measurements were performed using the built-in angle measurement tool in OsiriX in the sagittal plane since the degree of rotation required aligning the long axis of the glenoid to the 12-o’clock to 6-o’clock axis. The amount of rotation was defined as the rotation angle.15
Similarly, as described by Gross and colleagues,15 the sagittal en face view was divided via 5 cuts, throughout a superimposed best-fit circle that closely represents the glenoid.9,15,20 For both the UNCORR and CORR, glenoid width (AP distance) was measured on the axial image at the widest point from AP cortex across the glenoid face.
Utilizing the en face 3-D CT reconstruction view of the glenoid as the gold standard, patients were placed into 1 of 3 groups according to the degree of anterior GBL measured via the surface method.9,20 The groups were as follows:
I. 10% to 14.9% (N = 12)
II. 15% to 19.9% (N = 16)
III. >20% (N = 15)
Paired t-tests were used to compare all measurements between CORR and UNCORR scans for each of the 5 cuts. A P-value of .05 was used as the threshold for statistical significance in 2-tailed comparisons. Mean and standard errors are presented with standard deviations throughout the study. For interobserver reliability, the measurements between the observers, the intraclass correlation coefficient was calculated. All statistics were performed with SPSS (Version 22).
The study cohort was comprised of 19 left shoulders (44%) and 24 right shoulders (56%), including 36 male patients (84%) and 7 female patients (16%). The average age was 27.8 years (range, 21-40 years). The variability in measured difference, with respect to AP width, was 1.05 mm. The UNCORR CT scans required a mean correction for coronal inclination of 7.0° ± 5.8° (range, -8°-6°). The UNCORR CT scans required a mean correction for scapular angle of 30.2° ± 8.0° (range, 15°-49°). The mean angle of sagittal rotation required to align the glenoid face with the 12-o’clock to 6-o’clock axis was 24.2° ± 5.1 ° (range, 13°-30°). These results are summarized in Table 1. For all measurements, the intraclass correlation coefficient for independent observers for all cuts within the 3 groups was r >.900 in all cases.
On an optimized CT scan, over 5 standardized cuts across a best-fit circle of the inferior glenoid, there was a statistically significant absolute mean difference of 12.6% in axial AP glenoid width (2.86 mm ± 2.00 mm, P =.016) when compared with the UNCORR scan. This corresponds to a 3% to 21% error in measurement of the AP width of the glenoid.
For the entire cohort of 43 patients, the UNCORR scans underestimated the axial AP width (and thus overestimated GBL) in cut 1 (P =.003), and overestimated the axial AP width (and thus underestimated GBL) in cuts 3 to 5 (P < .001 for all) compared with that of the CORR scans. There was no significant difference between the UNCORR and CORR scans in cut 2 (P = .331).
For groups I (10%-14.9% GBL) and III (>20% GBL), the UNCORR scans underestimated the axial AP width (and thus overestimated anterior GBL) in cuts 1 and 2, while in cuts 3 to 5, the axial AP width was overestimated (GBL was underestimated) (Tables 2, 3). In Group II (15%-19.9% GBL), the axial AP width was underestimated (GBL was overestimated), while in cuts 2 to 5, the axial AP width was overestimated (GBL was underestimated). Overall, AP glenoid width was consistently underestimated in cut 1, the most caudal cut, while AP glenoid width was consistently overestimated in cuts 3 to 5, the more cephalad cuts.
The principle findings of this study demonstrate that UNCORR conventional 2-D CT scans inaccurately estimate glenoid width as well as inaccurately quantify the degree of anterior GBL. Underestimations of GBL may lead to insufficient treatment of clinically meaningful GBL, thereby increasing the risk of instability recurrence; whereas overestimations of GBL may lead to unnecessary treatment, subjecting patients to increased surgical morbidity. Therefore, the authors recommend correcting the orientation of the scapula in cases wherein clinical decisions are entirely based on 2-D CT, or using alternative methods for quantifying GBL, specifically in the form of 3-D reconstructions.
The use of axial imaging, with CT scans and/or magnetic resonance imaging, is growing in popularity for evaluation of both glenoid anatomy and GBL. Nevertheless, despite our improved ability to critically evaluate the glenoid using these advanced imaging modalities, the images themselves require scrutiny by clinicians to determine if the images accurately depict the true anatomy of the glenoid. As demonstrated by Gross and colleagues,15 conventional 2D CT scan protocols are not optimized to the anatomy of the glenohumeral joint, even in patients without GBL. Due to the alignment of the image relative to the plane of the scapula as opposed to the plane of the glenoid, UNCORR scans result in significantly different measurements of glenoid version (2.0° ± 0.1°) and AP glenoid width (1.2 mm ± 0.42 mm) compared with corrected scans, requiring an average 20.1° ± 1.2° of correction to align the sagittal plane. In the present study involving the patients with GBL, we also found that conventional, UNCORR 2-D CT scan protocols inaccurately estimate glenoid width and the degree of anterior GBL. In particular, AP glenoid width was consistently underestimated in the more caudal cuts, while AP glenoid width was consistently overestimated in the more cephalad cuts. Thus, anterior GBL was overestimated (AP glenoid width was underestimated) in the more caudal cuts, whereas anterior GBL was underestimated in the more cranial cuts (AP glenoid width was overestimated). Given that approximately 1 mm of glenoid bone corresponds to approximately 4% of glenoid width,16 even subtle differences in the interpretation of GBL may lead to gross overestimation/underestimation of bone loss, with significant clinical implications.
In the anterior instability patient population, clinical decision-making is often based on the degree of GBL as determined by advanced imaging modalities. In addition to other patient-specific factors, including age, gender, activity level, type of sport, and number of prior dislocations and/or prior surgeries, the quantity of GBL will often determine which surgical procedure needs to be performed. Typically, patients with >20% to 25% anterior GBL are indicated for a glenoid reconstruction procedure, most commonly via the Latarjet procedure (coracoid transfer).21-27 The Latarjet procedure remains an excellent technique for appropriately indicated patients, with historically good clinical outcomes and low recurrence rates. Complications associated with the Latarjet procedure, however, are not uncommon, including devastating neuropraxia of the axillary and musculocutaneous nerves, and occasionally permanent neurologic deficits.28 Thus, it is critical to avoid overtreating patients with recurrent instability and GBL. As demonstrated by this study, depending on the cranial-to-caudal location on the glenoid, current 2-D CT techniques may underestimate AP glenoid width, resulting in an overestimation of GBL, potentially leading to the decision to proceed with glenoid bone reconstruction when such a procedure is not required. On the contrary, overestimation of AP glenoid width, which occurs in the more cephalad cuts of the glenoid, is perhaps more worrisome, as the resulting underestimation of GBL may lead to inadequate treatment of patients with recurrent instability. Certainly, one of the main risk factors for failed soft tissue shoulder stabilization is a failure to address GBL. If clinical decisions are made based on UNCORR 2-D CT scans, which are often inaccurate with respect to AP glenoid width by an average 2.86 mm ± 2.00 mm (equivalent to 12.6% ± 6.9% GBL) as determined in this study, patients who truly require osseous glenoid reconstructions may be indicated for only soft tissue stabilization, based on the underestimation of GBL.
The current gold standard for GBL measurement is a perfect-fit circle performed on a 3-D CT scan.22 To that end, it would have been useful to measure the glenoids from this study on 3-D CT scans and compare the data with both UNCORR and CORR measurements. This would have provided a better understanding to what extent the CORR measurements on 2-D scans are relatable with the gold standard. As 3-D CT scans provide a better en face view of the glenoid, more accurate GBL measurements, and ease of 3-D manipulation, they have become more widely used across the country.29,30 Nevertheless, in situations where 3-D imaging is more challenging to obtain because of technology or cost limitations, having a strategy for ensuring proper orientation of 2-D scans would have a substantial impact on clinical decision-making. If such corrections are not made, the inaccuracy of current 2-D scanning protocols justifies the cost 3-D reconstruction protocols. The difference in GBL measurements are critical in cases of increasingly large degrees of GBL, as in these instances, the inferior glenoid becomes more of an inverted-pear shape as opposed to a perfect circle, and differences in CORR and UNCORR images are likely to be more profound.
This study has limitations, such as the relatively small sample size and the selection bias by the reviewers with potential differences in interobserver reliability. Further, minor modifications during the reformatting process may be found with each attempt to manipulate the images and may result in minor, insignificant differences in AP width measurements. Performing 1 or more additional CT scans on the same cohort of patients would have been helpful; however, due to the increased risk of radiation exposure, this was not performed. Performing CT scans on cadaveric specimens with GBL and applying the study methodology would also have been helpful to provide independent verification of our clinical findings; however, specimens were not available for this study. Another limitation of this study is that we did not compare our findings with the findings of glenoid width, and bone loss, as determined using the circle method, which is commonly utilized when 3-D reconstructions are available. In this study, the purpose was to utilize only the 2-D reformatted images, with the assumption that 3-D reconstructions are not always available, and cannot always be measured. To minimize selection bias, the investigators measured the correction effects within groups of patients with similar degrees of GBL (10%-14.9%, 15%-19.9%, and >20%). In addition, not all the selected patients showed degenerative glenoid changes or irregular glenoid shape indicating previous bone augmentation.
UNCORR 2D CT scans inaccurately estimate glenoid width and the degree of anterior GBL. The clinical implications of these findings are profound and suggest corrected 2D CT scans or 3D reconstruction allow measurements to be taken in the axis of the glenoid to accurately define the anatomy and quantity of anterior GBL in patients with shoulder instability.