3.1 Preoperative planning

Preoperative planning for total shoulder arthroplasty is a critical step to ensure proper final component placement. Numerous factors, such as glenoid bony wear, a patient's unique anatomic bony landmarks, as well as surrounding soft tissue and rotator cuff integrity, all play a role in determining the type of surgery performed and how it needs to be performed. Appropriate imaging is therefore paramount during the preoperative stage. Typically, plain radiographs can be taken to provide an overall assessment of bone quality and the joint, while CT scans provide more detailed data to allow better calculation of bone loss percentages, and magnetic resonance images (MRIs) can provide enhanced information regarding the surrounding soft tissue.

3.2 A plain radiograph versus a CT scan

Glenoid component positioning is crucial for TSA and RSA success. Historically, physicians have used plain radiographs to preoperatively appreciate glenoid anatomy, degree of version, and inclination. Antero-posterior views are used to appreciate the degree of bony arthritis, but can also elucidate osteophytes, humeral head migration, and ‘acetabularization’ of the acromion. The glenoid version, more specifically, is appreciated on axillary plain radiographs. The glenoid version was defined in 1992 by Friedman et al. as the angle between the glenoid line and the line perpendicular to the scapular axis.¹³ However, such measurements can become difficult with severe bony pathology, distortion from bony overgrowth, and suboptimal patient positioning. Using plain radiographs, Nyffeler et al. found that retroversion was overestimated in 86 % of measurements when compared to CT scans.¹⁴ Radiographic technique greatly influences imaging, with small variations in beam angle and tremendous differences in measurements.¹⁴

3.3 2D versus 3D CT imaging

Although CT scans capture images within a 3D plane, images are reported as 2D images for the viewer. Newer software makes it possible to reconfigure these images into a 3D construct for improved spatial visualization of relationships in the bony anatomy. The superiority of 3D CT scans compared to 2D CT scans has been well described. Scalise et al. compared 3D CT to 2D CT scans and found statistically significant differences in measured glenoid version and posterior glenoid bone loss.¹⁵ Other studies have found a statistically significant difference in glenoid version measurement between the two modalities when comparing individual measurements as well.^16–19 Differences in measurements of 5° or more are commonly found between the two modalities.^20,21 Improved visualization of the glenoid using 3D reconstruction eliminates the influence of positioning and alignment of the scapular plane. Systems using 2D imaging typically use the scapular plane for orientation, while 3D systems estimate the version using the relationship between the glenoid midpoint and the scapular plane, which could explain some of the variance between measurements.²¹

Pre-operative planning can influence surgical decision making as well. Scalise et al. found that surgical decisions were changed in 37 out of 96 cases when exposed to 3D imaging after 2D imaging.¹⁵ Appreciating a 3D reconstruction of the patient's glenoid morphology can guide the selection of implant, reaming depth, and degree of version correction. Overcorrection of the version with excessive reaming of the anterior side to neutral is known to increase the risk of humeral head subluxation and eventual glenoid loosening due to eccentric loading from medialization of the joint.²² Additionally, Ritter et al. showed that bone density measurements utilizing preoperative 3D CTs were good to excellent for predicting intraoperative bone quality.²³ Thirty cadaveric humeri were evaluated, and a machine learning model (support vector machine) was used to predict intraoperative bone density. The support vector machine algorithm demonstrated an accuracy of nearly 88 % in predicting intraoperative bone quality, with an area under the receiver operator characteristic curve of 0.93. The authors concluded that preoperative 3D CTs may provide surgeons with an additional tool in selecting appropriate patients for stemless humeral components during shoulder arthroplasty.²³

Despite many reports showing the advantages of 3D planning specifically for glenoid version measurement, Boileau et al. did not find a statistically significant difference between 2D and 3D measurements.²⁴ The clinical significance of these differences between 2D and 3D measurements is therefore debated. A systematic review concluded that although 3D imaging is advantageous for anatomic assessments, the benefit over 2D imaging only supports the use of CT in shoulder arthroplasty in general, regardless of the type used.²⁵ Further assessment of patient outcomes following pre-operative planning and the cost differences between each modality is important to further clarify the role of 3D planning in the future of shoulder arthroplasty.

3.4 MRI versus CT scans

Magnetic resonance imaging (MRI) is used for superior visualization of soft tissue pathology. A major indication for MRI use prior to shoulder arthroplasty is to clarify the presence of rotator cuff arthropathy, as this can warrant changing from an anatomic to a reversed construct. In patients who have available MRI imaging, it raises the question of whether obtaining a CT is warranted. The increased radiation exposure and cost of CT must be measured against its superiority for presurgical planning compared to MRI. However, there is evidence to suggest that low-dose CT protocols do not compromise image quality or preoperative surgical planning in shoulder arthroplasty and should be considered preoperatively.²⁶ Bohonos et al. found that CT scans were inferior to MRI for predicting glenoid width and equal to MRI for measuring baseplate central screw length.²⁷ However, 3D CT has been shown to be significantly more accurate for predicting glenoid bone loss compared to MRI and plain radiographs.⁷ Rosenthal et al. generated 3D reconstructions using either MRI or CT scans to assess glenoid wear in osteoarthritis shoulders.²⁸ A total of 29 shoulders were analyzed using either method, and two reviewers independently evaluated glenoid morphology and erosion. The authors concluded that 3D MRI was comparable to 3D CT with respect to axial glenoid bone loss, but coronal bone loss estimation was best evaluated by 3D CT.²⁸ Therefore, CT scans remain the gold standard for coronal glenoid bone loss but can be used in conjunction with other imaging modalities.²⁸

3.5 Assessment of osteoarthritis

Assessment of glenoid morphology and wear has historically been done using axillary radiographs.²⁹ Alternative classification systems for glenoid morphology have been described using CT imaging and 3D reconstruction.^30,31 The Walch classification system is the most commonly used to describe glenohumeral osteoarthritis.³⁰ It was originally described using CT imaging to define three types of glenoid morphology (types A, B, and C) and their associated glenoid wear patterns. Classification of patient glenoid morphology is necessary for surgical and intraoperative decision-making as it can alter arthroplasty outcomes.^29,32,33 Consistency in identifying wear patterns between imaging modalities has yielded contradictory results. Some investigators find overall intra-observer consistency to be slightly higher with radiographs compared to CT, while others find CT to be more consistent.^32,34 Scalise et al. found that measurements of posterior glenoid bone loss were significantly different when using 2D CT compared to 3D images.¹⁵ Observer agreement is also shown to significantly increase with the use of 3D compared to 2D imaging.¹⁵ Observer experience has a large influence on the identification of glenoid morphology, which advocates for the use of CT for more consistent recognition of wear patterns.³²

As shoulder osteoarthritis progresses, subchondral bone cysts commonly develop.³⁵ The presence of subchondral bone cysts may have implications for bone stock for potential shoulder arthroplasty candidates.³⁵ Pucchio et al. described their comparative study examining four segmentation techniques for measuring subchondral bone cysts in standard CT scans of patients with glenohumeral osteoarthritis.³⁵ Cyst volumes were manually measured using a micro-CT and served as the gold standard. Cyst volumes measured using qualitative and edge detection techniques had the greatest overall agreement compared to manual measurements using a micro-CT. The authors reported that these techniques demonstrated good to excellent reproducibility between graders and may provide utility for monitoring disease progression and aiding in preoperative planning.³⁵

3.6 Intraoperative considerations

The use of CT for shoulder arthroplasty is of particular interest for optimizing intraoperative decision making. The prediction of the method of arthroplasty (RSA vs. TSA) was found to be correct in 97 % of cases preoperatively in a single-surgeon study.³⁶ The majority of these errors were due to the identification of rotator cuff pathology rather than issues with bony anatomy. Preoperative planning with 3D CT scans increases the accuracy of glenoid implant size prediction, with prediction accuracy ranging from 81 to 86 %, with 100 % of incorrectly predicted implants being within 1 size of the implanted implant.^36,37 A 3D CT has also been shown to reasonably predict humeral stem implant size in both reverse and anatomic TSA. Wittman et al. demonstrated that when 3D CT was used preoperatively, stem size was found to be within the range of one adjacent size in nearly 88 % of the cases.³⁸ Werner et al. found observer agreement in choice of implant for 45 out of 50 cases when shown 2D CT images. A total of seven cases were revised after being shown 3D images for improved glenoid implantation.¹⁷ Freehill et al. showed that the use of 3D CT templating software resulted in glenoid sizes that were within one size of the actual implant 99 % of the time and an exact match 89 % of the time.³⁹ The authors also showed that for stemmed humeral prostheses, 88 % of cases were within one size of the preoperative template, and 83 % were a perfect match.³⁹ Increased accuracy in prediction not only decreases operative time and anesthesia time, but also has the potential to optimize surgical room supplies and the overall cost of equipment.³⁶

Gregory et al. published their findings on the use of preoperative CT scans for surgical templating of anatomic TSAs.⁴⁰ The authors evaluated the bone stock of proximal humeri by averaging the Hounsfield units (HU) across anatomically defined regions. They found that preoperative metaphyseal cancellous bone density was associated with intraoperative conversion to a stemmed prosthesis in comparison to those who were able to undergo a stemless implant (5.5 ± 11.2 HU for stemmed components versus 47.6 ± 29.4 HU for stemless components, P < 0.001).⁴⁰ All patients who were intraoperatively converted to a stemmed prosthesis had a preoperative CT with a metaphyseal cancellous bone density less than 20 HU.⁴⁰ Therefore, the authors report that a threshold of 20 HU can be used to predict which patients would likely benefit from a stemmed prosthesis when undergoing anatomic TSA.⁴⁰ Similarly, Levin et al. described their CT measurements used to predict if surgeons would intraoperatively convert from a stemless to a short-stemmed humeral prosthesis during anatomic TSA.⁴¹ The authors reported a HU cutoff of 14.4 that resulted in a 95 % sensitivity and 100 % specificity for intraoperatively using a stemmed humerus prosthesis.⁴¹ The use of preoperative CT shows promise in aiding surgeons in their humeral fixation strategy prior to entering the operating room.⁴¹

Interestingly, Almeida et al. explored baseplate positioning in RSA cases where no intraoperative 3D technology was used to see if the parameters were within an acceptable range. They found that baseplate positioning was overall within acceptable range; however, one out of three of the baseplates was positioned outside of the recommended version parameters, and one out of three of the upper screws did not reach the base of the coracoid.⁴²

3.7 Patient-specific instrumentation (PSI)

The construction of PSI using CT-based preoperative planning improves implant selection and reduces the risk of glenoid malpositioning in both TSA and RSA.⁴³ Berhouet et al. compared measurements of the glenoid in surgeons who have access to 3D preoperative planning technologies to those who only had access to the surface of the glenoid visualized during surgery and found significantly more accurate positioning of the RSA glenoid component, supporting the use of PSI and 3D planning.⁴⁴ Iannotti et al. found the accuracy of guide pin positioning increased by 4.5 ± 1.0° in version, 3.3 ± 1.3° in inclination, and 0.4 ± 0.2 mm in location (all P values were significant) when comparing the use of PSI with CT planning to standard instrumentation for patients undergoing both TSA and RSA.⁴⁵ PSI has been shown to increase precision and reliability in guide pin insertion in vitro.⁴⁶

Suero et al. compared postoperative glenoid component positioning in TSA and RSA cases using PSI guides to CT-based preoperative plan positioning. The absolute difference in glenoid positioning was found to be 3.4 mm.⁴⁷ Overall, quantitative analysis showed alignment between the preoperative plan and postoperative component placement using PSI guides.⁴⁷ Because this study was limited by the lack of a control group, the findings can only support the feasibility of using PSI guides and cannot accurately denote whether this method of implantation is superior. Similarly, Kwak et al. compared 20 conventional RSAs versus 19 CT-based RSAs using PSI.⁴⁸ The authors calculated the difference between preoperative targets and postoperative measurements for screw length and angle, baseplate version, inclination, translation, and rotation. The authors noted that the mean differences in screw lengths and angles were significantly smaller in the PSI group and that the mean difference in baseplate rotation was significantly lower in the PSI group as well (4.5 versus 10.6°; P < 0.001).⁴⁸ The authors also noted that 10 screws violated the spinoglenoid notch in the conventional group as opposed to only 2 screws in the PSI group (P = .014).⁴⁸ They concluded that PSI was able to improve the reproducibility of their operative plan and potentially reduce the risk of neurovascular injury in RSA by more frequently avoiding the spino-glenoid notch during screw placement.⁴⁸

Additionally, Rojas et al. demonstrated in their cadaveric RSA study of twenty specimens that PSI had fewer outliers between planned and post-osteotomy humeral retrotorsion and height, relative to standard cutting guides.⁴⁹ Outliers were defined as > 5° inclination, >10° retrotorsion, and >3 mm height.⁴⁹ There were eight outliers in the standard cutting guide cohort, with only one outlier in the PSI cohort.⁴⁹ The authors noted that most outliers in the standard gutting guide group were due to >10° of humeral retrotorsion.⁴⁹

The PSI techniques have also been used in shoulder arthroplasty to help address bone loss.⁵⁰ Karpyshyn et al. described their novel surgical technique using custom 3D-printed cutting guides to create a patient-specific bone graft in the setting of an RSA glenoid deformity.⁵⁰ Through the use of 3D-printed custom cutting guides, surgeons can prepare the bone graft and accurately correct glenoid deformity with the use of a single guide, as opposed to using a freehand technique and using standard cutting guides that cut the graft at standard angles.⁵⁰ Sadeghi et al. also published their work evaluating PSI for guide pin positioning in RSA in glenoid deficiency.⁵¹ There were 150 scapula models used, and post-operative CT scans of the samples with guide pins were obtained to assess the accuracy of guide pin position compared to preoperative CT templating.⁵¹ The authors concluded that using CT-based PSI guides showed advantages in glenoid component positioning over freehand and conventional non-PSI guides, but further investigation into clinical outcomes is necessary.⁵¹

Important drawbacks to PSI include the associated cost, the increased time spent for preoperative planning, and manufacturing.⁴⁶ However, there are examples of in-house 3D-printed cutting guides, which reportedly take less than 5 h and $30 to produce.⁵² Although not all surgeons have access to 3D-printed materials to be used intraoperatively, PSI holds promise for continuing to improve its affordability and accessibility.

The current investigation into whether PSI improves clinical outcomes is ongoing. Hwang et al. published their retrospective matched cohort analysis from a multicenter prospective cohort of patients undergoing RSA with preoperative 3D CT planning and a minimum of two-year follow-up.⁵³ The patients were placed into two cohorts contingent if a standard manufacturing guide or PSI was used for glenoid guide pin placement. A total of 56 patients were in the standard guide cohort, and 122 underwent PSI.⁵³ The authors found that both groups had similar improvements in PROs regardless of PSI being used, but patients who received PSI had greater improvement in postoperative strength.⁵³ Further work on the clinical impact of patients receiving PSI for shoulder arthroplasty continues.

3.8 Navigation

Navigation has revolutionized the field of arthroplasty, specifically in the knee and hip. However, navigation on the shoulders has recently been more up and coming. Within the last decade, studies have been conducted to examine the benefits of navigation in shoulder arthroplasty. Similar to preoperative planning software, navigation systems use preoperative CT imaging for both planning and intraoperative guidance. In 2015, Venne et al. compared CT-navigated and conventional screw placement in RSA cases performed by five surgeons at a single institution. Although the screw entry point and accuracy of the inferior screw had no significant difference based on approach, end point, and angulation, the three screws showed significant improvement in navigated cases..⁵⁴ Conversely, another study found that axial screw angulation of the anterior and inferior screws was significantly more posterior with navigation compared to manual arthroplasty, while no differences were seen with the superior and posterior screws.⁵⁵ Screw purchase length in anterior and posterior screws was found to be significantly longer (at least 6 mm) with navigation compared to manual arthroplasty, while superior and inferior screws had no significant differences.⁵⁵ Velasquez et al. published their meta-analysis of navigation's impact on baseplate screw configuration in RSA, which included 633 shoulder arthroplasties and six trials.⁵⁶ The authors concluded that intraoperative navigation significantly improves baseplate screw placement, which tends to improve screw purchase length and the overall use of fewer screws to achieve glenoid fixation.⁵⁶

Due to concern over removing too much bone from the glenoid, physicians often under-ream the glenoid, leading to unintentional excessive retroversion.^57,58 Intraoperative navigation using uploaded CT imaging can significantly improve version angle.⁵⁷ Larose et al. performed a retrospective review of all shoulder arthroplasties (TSA and RSA) implanted using a single computer navigation shoulder system utilizing preoperative CT-based software.⁵⁹ A total of 16,723 shoulder arthroplasties were reviewed.⁵⁹ In 98 % of cases, the navigation system was used for the entirety of the case.⁵⁹ There were nine reported coracoid fractures (0.05 %) and minimal differences between preoperative planning and intraoperative execution regarding inclination (0.2° ± 2.04°), version (0.6° ± 1.96°), and glenoid pin placement (1.90 ± 1.2 mm).⁵⁹ Further assessment of the long-term survivability of these implants will be necessary, but given the current literature, the safety and accuracy of navigation hold promise.⁵⁹

Navigated augmented reality (AR) has also been investigated to assess its accuracy in glenoid component placement in shoulder arthroplasty.^60,61 Rojas et al. utilized preoperative 3D CT scans on 12 cadaveric shoulders and used a head-mounted display to navigate glenoid component inclination, retroversion, depth, and rotation.⁶² Postoperative CTs were obtained and compared to preoperative CTs. The difference between planned and postoperative glenoid component positions is as follows: 1.0° ± 0.7° for inclination, 1.8° ± 1.3° for retroversion, 0.7 ± 0.6 mm for depth, and 1.7° ± 1.6° for rotation.⁶² The authors concluded that the use of an AR system led to low deviations between plan and postoperative glenoid component position.⁶²