CORR Insights®: Patient Posture Affects Simulated ROM in Reverse Total Shoulder Arthroplasty: A Modeling Study Using Preoperative Planning Software

Where Are We Now?

When total shoulder arthroplasty was first developed, orthopaedic surgeons used it to treat nearly every shoulder condition, although our understanding of the biomechanics of this procedure at that time was rudimentary at best. Biomechanical studies from that era set out to determine how well shoulder arthroplasty replicated normal shoulder anatomy, and the goal of the procedure was to replicate normal shoulder biomechanics as much as possible. Shoulder arthroplasty has come a long way since Neer introduced devices we would recognize as reasonably contemporary some 50 years ago [11, 12].

The introduction of reverse total shoulder arthroplasty (rTSA) changed many of our assumptions about how the shoulder works and would ultimately lead to numerous biomechanical concepts and findings about this complicated joint. Indeed, rTSA opened the door for the successful treatment of many shoulder conditions that were not amenable to treatment with anatomic total shoulder arthroplasty. These conditions presented a wide variety of challenges, not only for treatment, but to our understanding of the reasons why some shoulder arthroplasties succeeded and others did not.

Of the many laboratory approaches that have helped us learn more about orthopaedic biomechanics, I think computer modeling has been perhaps the most helpful in terms of developing contemporary approaches to shoulder arthroplasty. Today, we can perform a CT scan of the shoulder complex and determine the morphology of the glenoid and allow a better assessment of the glenoid shape than with plain radiographs alone. Computer modeling helps us in several important ways. It can create guides to direct pin placement so that the glenoid can be reamed to provide the most support for the glenoid components, it can assist us in determining where the glenoid bone should be removed for augmented glenoid components, it can be used for the creation of custom glenoid components for larger glenoid defects [1-3, 13], and finally, computer modeling helps us determine the best position of the glenoid and humeral components so that ROM of the shoulder can be improved with the implants in place.

Interestingly, when the goal of the biomechanical models is to modify the implant position to improve shoulder motion, the model must include not only prosthesis positioning but also prosthesis design variables thought to influence shoulder motion. For rTSA, these variables include humeral tray designs such as onlay versus inlay, glenosphere center of rotation and offset, and baseplate position and inclination [6].

The current study by Moroder et al. [10] adds to the sophistication of computer modeling of the shoulder and of rTSA by incorporating scapulothoracic relationships into the model. This idea pays homage to Dr. Ben Kibler, who reminded us that the glenohumeral joint does not exist alone, but that its relationship to the scapula and thorax must be respected [7].

Moroder et al. [10] have made a Herculean attempt to take computer modeling of rTSA to a higher level by evaluating the effects of scapular position on the subsequent motion of rTSA. In this study, the authors divided the posture of the torso of 30 cadavers into three groups described as upright posture with a retracted scapula, an intermediate group, and a third group defined as kyphotic with a protracted scapula. The scapular internal rotation was the defining parameter separating the groups. The authors then implanted a short stem humeral component into the model rTSA but modified the amount of retroversion of the component. They also varied the head-neck angle of the humeral tray to the stem as 135°, 145°, or 155°. They did not vary the implants by inlay or onlay humeral tray configurations. On the glenoid side, they varied the humeral sphere diameter to include 36-mm concentric, 36-mm eccentric, and 42-mm concentric spheres. The variables tested created more than 3700 configurations, which could be studied with the computer model to determine the maximum ROM of the humerus before there was impingement upon the scapula. They found that for those torsos with a kyphotic posture, it was advantageous to have the humeral component in more retroversion, a lower neck-shaft angle of the humeral tray, and a larger glenosphere.

Where Do We Need to Go?

Most biomechanical models of shoulder arthroplasty make several assumptions because they cannot typically account for all of the static and dynamic variables that influence the final motion of the implants. This is particularly true in the shoulder, where the interaction of the humerus to the glenoid, the motion of the scapula on the thorax, and the muscle activity all contribute to the final motion. These issues are apparent in this study; the scapular position was static, and the effect of scapulothoracic motion upon the interaction of the humerus to the scapula in different scapulothoracic positions was not studied. Similarly, there are no soft tissue parameters included in this model, especially the impact of soft tissue contracture as well as muscle activity upon impingement. My personal bias is that while implant position is important for clinical outcome, the major factor impacting final ROM is the degree of soft tissue contracture prior to surgery.

Our measurements should also be considered a limitation of biomechanical models of shoulder arthroplasty. Numerous assumptions have to be made in creating the software, which may influence the results. In this study, the software is proprietary and was “modified,” but that makes it difficult to compare it to other biomechanical models or to real-life shoulder motions. Another limitation of biomechanical models is that typically only one implant system is studied, and extrapolation to other implant designs cannot be done.

Lastly, the relationship of the biomechanical models of shoulder arthroplasty to final clinical results has remained elusive. There is a lot of “slop” in the shoulder, meaning that the shoulder is not a highly constrained joint. As a result, the soft tissues greatly contribute to shoulder mechanics, more so than in other joints. In the shoulder, there is high variability in the shape of the glenoid, the configuration of the proximal humerus, the laxity or contracture of the soft tissues, the presence or absence of large osteophytes, and the presence or absence of the rotator cuff. When using computer-driven guides for the central pin of the glenoid prior to reaming, a deviation of a few degrees and millimeters from the modeled position was postulated to be acceptable, although it was not known if that amount of inaccuracy really impacts the final clinical result [5]. As of this writing, no study has demonstrated the clinical efficacy of computer modeling in its clinical results, including ROM, scapular notching, or implant loosening.

My observation is that for most patients, the main clinical objective is pain relief, whereas ROM is a secondary objective. Fortunately, pain relief occurs in most patients regardless of their final ROM, and often it has little to do with component position alone. Shoulder arthroplasty is primarily a pain-relieving operation, and it may be that only specific motions relate to final outcomes. Future studies using computer navigation should prove how the calculations relate to patient satisfaction and clinical results. For instance, ROM could be less important than other variables that drive clinical success, such as weakness, depression, anxiety, and other patient-reported outcome measures. We need more studies to fill this gap in our knowledge before these models can be reliably used in clinical practice.

The current gaps in our knowledge need to be pursued before the clinical relevance can be ascertained. The computer models currently available on the market are proprietary, and while they are available on a wide scale, the data for determining its accuracy and its validity are not available to other basic scientists in the field. Transparency for the models and the assumptions upon which they are based need to be shared between researchers.

How Do We Get There?

The utility of biomechanical studies in guiding treatment of patients undergoing rTSA depends on the accuracy of the models as they relate to clinical practice and thus to clinical result. A previous study [4] suggested that the modeling performed by some examiners has low reproducibility and interexaminer or intraexaminer reliability. Future studies should address whether the variability of the results between examiners has any clinical significance. It may be that just estimating the best position of the components produces acceptable clinical results.

Another major hurdle for these biomechanical models is that they cannot account for soft tissue constraints, which influence shoulder motion. Understanding the interactions of the capsular contractures and how they affect final ROM would be a major step forward in predicting shoulder motion. There needs to be a way to not only determine capsular length but add it to the biomechanical models. When measuring shoulder ROM clinically, it has been shown that glenohumeral motion can be isolated from combined glenohumeral and scapulothoracic motion [9]. The authors here suggest that there are three basic types of posture and that they should be considered in preoperative planning, but this defies the complexity of the shoulder and has a way to go before it can be applied to the clinical situation.

Once the biomechanical models have been validated to include glenohumeral and scapulothoracic motions in a particular patient, there is an opportunity to determine which component positions can result in the best ROM for function and to avoid notching. In the shoulder, small changes of a millimeter or two can have important impact upon the conclusions of a biomechanical model, and this needs to be determined for any and all implant systems. Finally, the predicted ROM by the biomechanical models and those measured clinically should be compared to determine if the models predict final motion accurately or not.

We also must determine a hierarchy of variables according to their biomechanical and clinical importance. A clinical study of 5774 shoulder arthroplasties found that 19 demographic and clinical variables could be used to predict patient outcomes [8]. The relationship between computer modeling and clinical result requires an easy-to-perform model. The data produced will need to become part of the assessment of clinical and radiological results over time. Biomechanical models of this nature have the potential to be a helpful adjunct to determining patient result and satisfaction, but future studies needs to determine which variables are most important and which lead to patient satisfaction and longevity of the replacement.