4.13 Reverse geometry replacement

Conventional shoulder arthroplasty has been extremely successful in many forms of arthritis. The aim is to resurface the arthritic joint, but also to restore joint biomechanics. Deficiency of the soft tissues, and in particular of the rotator cuff, is therefore a major problem. Contractures can be released, the subscapularis can be lengthened, but an irreparable cuff tear is by definition an insurmountable obstacle to restoring normal biomechanics.

Whilst a tear is the most common cause of cuff dysfunction, there are other situations where the proximal humeral anatomy is distorted and the cuff is either physically or functionally compromised. These include certain fractures and fracture sequelae, as well as certain revision situations. In all of these cases poor active elevation and an unstable centre of rotation secondary to a deficient cuff cannot be corrected by a conventional shoulder prosthesis.

Neer recognized the challenge posed by arthrosis associated with massive irreparable cuff tears and suggested that it was realistic to pursue ‘limited goals rehabilitation’. The limited results of hemiarthoplasty in this setting have been confirmed by several other authors who reported inconsistent pain relief and small improvements in active elevation.

Conventional total shoulder arthroplasty fared no better, as superior eccentric loading of the glenoid component by the unstable humeral head increased the risk of premature loosening, the so-called rocking horse phenomenon.

In the early 1980s, a number of constrained shoulder prostheses were introduced. Increased constraint stabilized the centre of rotation, but at the expense of premature aseptic loosening, and most never progressed beyond the experimental. This same lesson had already been learned with excessively constrained prostheses in other joints.

In 1985, Grammont introduced a revolutionary design of semi-constrained reverse geometry total shoulder replacement (RSR). The idea of reversing the ball and the socket was not new. However, previous designs had sought to preserve the shoulder’s anatomical centre of rotation by using a glenoid component shaped like a chess pawn, with a small head offset from the glenoid base-plate by a neck. Grammont’s prosthesis, the Delta, used a large hemispherical glenoid component applied directly to the surface of the glenoid in order to minimize torsional stresses, yet stabilize the centre of rotation and optimize range of movement.

This design has progressed well beyond the experimental and its popularity has encouraged a number of imitations, each striving to ‘improve’ on the original design.

Shoulder elevation is a rotatory movement requiring a muscular moment to overcome the moment resulting from the weight of the arm. In practice, this is provided by the deltoid and the supraspinatus, but the initial forces produced by the deltoid when the arm is alongside the body are predominantly vertical shear. It is the role of the rotator cuff to oppose this upward force and stabilize the centre of rotation, in order to convert a vertical linear force into a rotatory movement. A useful analogy is of a person standing on the edge of a cliff holding a ladder on a rope that is secured to a rung halfway down its length. If they pull on the rope, the ladder will simply slide up the cliff. However, if they place their foot on the top rung of the ladder and then pull on the rope, the ladder will swing away from the cliff.

The situation is, however, a little more complex than this. The cuff does not simply oppose the vertical component of the deltoid, but also generates compressive forces that help to stabilize the centre of rotation in the sagittal as well as in the coronal plane. Furthermore, the deltoid itself also generates compressive forces as elevation increases, and beyond 60 degrees these compressive forces exceed its vertical shear force, thus diminishing the role of the cuff.

Post defined a massive rotator cuff tear as measuring 5cm or more in the sagittal plane. This is, however, a somewhat arbitrary figure, and a massive tear is perhaps more usefully defined as one involving at least two complete tendons. In practice, this means the supraspinatus plus the infraspinatus, and/or the subscapularis. Any remaining cuff is also likely to be degenerate, and its function may be further compromised by fatty degeneration of the muscle bellies.

Without a functional cuff, the unopposed upward force generated by the deltoid muscle causes the humeral head to migrate proximally and ultimately to abut the acromion. It is still a matter of some speculation why certain patients are able to maintain useful active elevation despite this biomechanical disadvantage, yet others are not and exhibit what has been termed pseudoparalysis. An intact coracoacromial arch should provide a fulcrum to stabilize the centre of rotation, so why do some patients lose active elevation? This may sometimes be explained by pain inhibition of the deltoid, yet pseudoparalysis is often painless. In these cases, upward migration of the humerus is probably sufficient to functionally lengthen and thus seriously weaken the deltoid. Furthermore, a massive cuff tear, particularly one involving subscapularis, may destabilize the centre of rotation of the shoulder in the sagittal as well as in the coronal plane. The humeral head may therefore sublux anteriorly as well as superiorly, thus diminishing the fulcrum provided by the acromion.

A massive cuff tear is often a gradual process of cuff wear rather than a sudden catastrophic failure of the cuff. This means that the bony surfaces have the opportunity to remodel, a process which can eventually lead to acetabularization of the glenohumeral joint (Figure 4.13.1). This adaptive process has been described by Hamada in five stages:

As mentioned earlier, previous attempts at RSR had used a glenoid component shaped like a chess pawn in order to restore the centre of rotation to its anatomical location (Figure 4.13.2A). This meant that the upward force of the deltoid was applied to the head of the glenoid component, and the neck of the component provided a lever arm. The end result was the application of excessive torsional forces to the glenoid base-plate causing early loosening. What is more, the small size of the glenoid head meant that impingement between the components limited the potential range of movement. Broström published the only series in the literature for this type of design (the Kessel prosthesis). Pain relief was achieved in 90%, but average active elevation was only 35 degrees and the reoperation rate was 26%.

Grammont was therefore facing a formidable challenge. His prosthesis needed to stabilize the centre of rotation in the absence of a functional cuff, restore a functionally weakened deltoid, optimize the range of movement allowed by the prosthesis, yet avoid premature loosening.

The crucial element of his design was a large hemispherical glenoid component applied to the glenoid without a neck (Figure 4.13.2B). This had the following consequences:

The use of a hemispherical component applied directly to the glenoid means that the centre of rotation is at the bone–prosthesis interface. There are therefore no torsional forces applied to this interface as there were with designs that used a neck to preserve a more anatomical (lateral) centre of rotation (Figure 4.13.2C).

The humeral component articulates under the glenoid hemisphere, which means that the humerus is displaced medially and distally relative to its normal position. In the normal shoulder the middle fibres of the deltoid course around the greater tuberosity, giving the shoulder its rounded contour. The tuberosity effectively acts as a cam, increasing the tension in the deltoid muscle as well as altering the angle at which it acts on the humerus. Medial displacement of the humerus removes this cam effect and the deltoid has a more linear course between the acromion and its humeral insertion, effectively lengthening the muscle. It is therefore vital that the humerus is also displaced distally, not only to restore the deltoid tension that was lost as a result of the deficient cuff (and the consequent upward migration of the humeral head), but also to compensate for this effective lengthening of the deltoid. Clinical measurements have demonstrated that mean lengthening of the upper limb following RSR is 1.5cm.

As has been well documented in hip arthroplasty, rotation around a small sphere is typically limited by impingement between the components. In the shoulder, scapular rotation can compensate to a degree for limited glenohumeral movement, but the use of a large glenoid hemisphere maximizes component excursion. Furthermore, in the absence of a neck on the glenoid component, a larger hemisphere also reduces the risk of impingement between the humeral cup and the inferolateral aspect of the scapula. This will be discussed later on the subject of scapular notching.

The most important factor in minimizing the risk of premature loosening of the glenoid is medialization of the centre of rotation and the attendant abolition of torsional forces at the bone–prosthesis interface. However, particularly during the earlier stages of elevation, the glenoid component is nevertheless subjected to significant shear forces that may also jeopardize glenoid fixation.

Grammont’s first design for the glenoid component was a monoblock which fitted over and around the glenoid like a barrel and was fixed with cement. However this was rapidly abandoned in favour of the current design which provides superior fixation in the limited bone stock of the scapula and greater resistance to shear forces. The glenoid component now has two parts: the metaglene and the glenosphere.

The metaglene is a circular baseplate with a central pillar and a hydroxyapatite coating. Its fixation to the glenoid is augmented by two divergent screws whose heads lock into the plate, as well as two non-locking screws. Not only is the bone-stock limited by the shape of the scapula, but in many of these elderly patients the bone itself is of poor quality. The locking plate configuration therefore provides superior pullout strength as compared to a central peg alone, or to non-locking screws. The superior screw is aimed at the base of the coracoid and the inferior one into the inferior pillar of the scapula as these represent the thickest areas of bone in the scapula medial to the glenoid.

The hemispherical glenosphere fits over the metaglene and is fixed with a peripheral morse taper and a central countersunk screw.

On the humeral side a polyethylene cup is carried by a modular humeral component. The latter is composed of a stem and an epiphyseal component which may be cemented or uncemented. There is, however, very little evidence to support the use of the uncemented stem at present.

The Delta RSR was originally designed for elderly patients with painful cuff tear arthrosis who had failed to respond to conservative measures. In essence this meant Hamada stages 3–5 (Figure 4.13.3). However, patients without bony changes (Hamada stage 2) who fail to respond to simpler surgical procedures such as arthroscopic debridement and biceps tenotomy may also benefit from RSR.

Whilst most authors agree that RSR should be reserved for patients over the age of 70 years, most series do include a number of patients who are younger than this. Those who still have active elevation above shoulder height despite cuff-tear arthrosis may obtain good results with a conventional hemiarthroplasty. In the presence of painful pseudoparalysis, the case for RSR in a patient under the age of 70 becomes stronger, but both surgeon and patient should be aware of the uncertain long-term results, and each case should be judged on its individual merits.

RSR is indicated for revision cases where there is either significant distortion of the proximal humeral bony anatomy or a deficient rotator cuff. The most common situation where this is encountered is in a fracture hemiarthroplasty that has failed because of tuberosity non-union or malunion.

Painful malunion or non-union of a three- or four-part proximal humeral fracture is another difficult surgical situation. Hemiarthroplasty alone cannot restore cuff function unless combined with osteotomy of the tuberosities, but the results of this are often disappointing and there is a high rate of tuberosity non-union. RSR offers a solution without the need for tuberosity osteotomy (Figure 4.13.4).

In the acute displaced four-part fracture, hemiarthroplasty remains the treatment of choice, but accurate and durable reconstruction of the tuberosities around the prosthesis can present a considerable challenge. Malposition, non-union, or migration of the tuberosities are significant risks and can seriously compromise results. RSR has been suggested as an alternative, particularly if the tuberosities are comminuted or osteoporotic as is often the case in the elderly.

Cuff tears have a high prevalence in the elderly, but few cause sufficient symptoms to warrant a RSR. Furthermore even the original RSR, the Delta, is still quite a new prosthesis. Most series are therefore relatively small and follow-up is at best short to medium term.

Only five series with more than 20 patients have been published for the Delta RSR to date (Table 4.13.1). As one might expect, the results for primary reverse arthroplasty are significantly better than for revision surgery. In the largest series to date (Wall et al. 2007), elevation improved from 76 degrees to 142 degrees in the CTA group, but from 58 degrees to 118 degrees in the revision group. Constant score improved from 27.7 to 65.1 in the former versus 19.7 to 52.2 in the latter.

In the only published review of acute three- or four-part fractures treated with RSR, mean active elevation was 97 degrees despite secondary displacement of the tuberosities in 53%. Mean age/sex adjusted Constant score was 66%. Unlike hemiarthroplasty, clinical results were not influenced by tuberosity healing.

Whilst RSR has been shown to restore active elevation, active external rotation is not significantly improved. The main determinant of postoperative active external rotation is therefore the preoperative condition of the posterior cuff. Latissimus dorsi transfer has been described in association with RSR. Active external rotation did not improve significantly with the arm at the side, but functional external rotation in the Constant score did. The authors argued that this resulted in a better ability to position the hand in space and thus superior function.

Survivorship analysis of a multicentre series of 80 cases demonstrated 10-year survival of 91% for revision and 84% for glenoid loosening, but only 58% for ‘absolute Constant score <30’ as an end point. Confidence intervals were, however, wide.

Reported complication rates range from 19–50%, but rates vary considerably between primary and revision cases. In the largest series to date, a complication rate of 13% was reported for primary surgery as opposed to 37% for revision cases. This large difference has been confirmed by others.

As well as the complications that affect conventional shoulder arthroplasty, there are three that appear to be a particular problem with the reverse design, namely notching, instability, and acromial fracture.

Whilst the absence of a neck on the glenoid component conveys great benefits, the reduced offset between the humerus and the scapula can also have a significant detrimental effect: impingement between the superomedial aspect of the humeral component and the lateral pillar of the scapula just below the glenoid. This commonly results in damage to the humeral cup and the appearance of a notch on the scapula. This has been reported in 44–68% of cases.

In many cases the notch is small and does not extend as far as the inferior screw, but it can extend to and even beyond this screw in up to 28% of cases (Figure 4.13.5). These larger notches are difficult to explain on the basis of physical impingement alone, and are probably due to osteolysis induced by polyethylene wear particles released from the damaged humeral cup.

The risk of notching may be reduced by inserting the glenoid component with a slight inferior overhang relative to the glenoid in order to minimize the physical impingement (Figure 4.13.3). This has been confirmed in a cadaveric study and also a retrospective clinical study.

Despite the high prevalence of notching, loosening of the glenoid component has not been a significant problem to date, even with very large notches. It should be noted however that there are no long-term reviews in the literature and notching is arguably the greatest threat to long-term survival of RSR. Because of this, most authors recommend that RSR should be reserved for patients over the age of 70 at present.

In the absence of a functional cuff, a semi-constrained reverse prosthesis depends on deltoid tension to maintain compression between the congruent joint surfaces. Inadequate tension can cause instability that may be addressed by increasing the thickness of the humeral cup or adding a neck extension under the cup in order to increase deltoid tension.

At the other end of the spectrum, excessive deltoid tension can cause a fracture of an acromion that may have been eroded and weakened preoperatively by the action of the humeral head (Figure 4.13.6). However, such fractures are often asymptomatic and may present as incidental radiological findings.

The Delta RSR has proved to be extremely popular as it offers a unique solution to a difficult orthopaedic problem. It has therefore encouraged a number of imitations with a variety of purported improvements. Several of these prostheses are very close to the Delta with small changes designed essentially to improve ease of implantation. Only time will tell whether any of these small changes have any unforeseen side effects.

There are, however, some new reverse designs that have moved further away from the Delta concept, mainly in an effort to reduce notching. The danger is that these changes may sacrifice the principles which have made the Delta RSR successful. For example, one proposed solution has been to use a glenoid component that is between a sphere and a hemisphere. This may reduce the risk of impingement between the humeral component and the scapular neck, but it does so by lateralizing the centre of rotation relative to the bone–prosthesis interface, in effect reintroducing a neck onto the glenoid component (Figure 4.13.3C). Although the offset is not as great as on older unsuccessful reverse designs, it remains to be seen whether it creates sufficient torsional forces at the glenoid component–bone interface to induce premature loosening.

The Delta itself needs longer follow-up in order to determine whether notching is a threat to its survival. Furthermore, whilst it has been shown retrospectively that a low position of the metaglene is associated with a reduced incidence of notching, prospective studies are required to confirm that this can be reliably achieved and that notching can indeed be avoided or at least minimized.

Modern RSR has proved itself as a useful tool in the treatment of a number of conditions where conventional shoulder arthroplasty offers limited and uncertain results because of deficient cuff function. It has been shown to provide reliable pain relief and also to restore active elevation, though active external rotation often remains limited. However, it is important to remember that only short- to medium-term results are available at present. Early loosening has not been a problem, but notching could potentially lead to failure in the longer term. Careful patient follow-up is therefore essential.