Preparation of the Specimens
Glenohumeral joints were obtained from fresh-frozen cadavera. Each shoulder was dissected to the level of the rotator cuff, with preservation of the distal humeral condyles and the site of insertion of the deltoid muscle. Shoulders that were found to be abnormal, on either radiographic or physical examination, were eliminated from the study. Seven specimens from the cadavera of individuals who were a mean of seventy-three years old (range, fifty-four to eighty-seven years old) at the time of death were used. Each joint was vented to eliminate any effects of intra-articular pressure. The joints were stored frozen, thawed at room temperature before testing, and kept moist with a protease-inhibitor solution during experimentation.
Simulated deltoid muscle forces were applied with strands of low-stretch braided Dacron, which were attached to the site of insertion of the deltoid on the humerus by a transosseous bolt. The strands of Dacron were placed through three eyelets on the scapula that corresponded approximately to the mid-points of the anatomical origins of the anterior, middle, and posterior portions of the deltoid muscle (the tip of the coracoid and the anterior and posterior portions of the acromion, respectively).
For simulation of the rotator-cuff muscles, Dacron cord was sewn, in a Bunnell suture pattern, directly into the tendinous insertion of the subscapularis, the supraspinatus, and the combined infraspinatusteres minor complex. The infraspinatus and subscapularis fossae were cleaned of all soft tissue, and the muscle centroids were marked. Pulleys were placed on the potting base so that the lines of action of the rotator cuff were aligned with the centroids marked on the muscles.
As the entire protocol was performed at 90 degrees of humeral elevation, the scapula was potted in Bondo (Dynatron/Bondo, Atlanta, Georgia) with its medial border aligned at 30 degrees of scapular elevation in order to achieve a 2:1 ratio of glenohumeral-to-scapulothoracic motion. To ensure proper alignment, the scapula was held in its desired position with alignment bolts and the Bondo was injected into the base with Monoject catheter-tip syringes (Sherwood Medical, St. Louis, Missouri). Muscle forces were simulated with a hand-crank system that has been described elsewhere16.
Technique for Arthroplasty
Each shoulder-implant system (DePuy, Warsaw, Indiana) consisted of a humeral stem, a humeral head, and a glenoid component. A titanium-alloy humeral stem of appropriate size (eight or ten millimeters in diameter) with a reverse locking taper fitting at its neck was chosen for each specimen. Six cobalt-chromium-alloy humeral head components, with a nominal radius of curvature ranging from twenty to twenty-five millimeters in one-millimeter increments, were used in each specimen. In order to maintain consistent capsular tensioning, all of the humeral head components had the same neck length, so that the distance from the resected surface of the humerus to the most proximal point of the head was constant. Keeled glenoid components were made of ultra-high molecular weight polyethylene, and the articular surface had a nominal radius of curvature of twenty-five millimeters. The glenoid component was combined with each of the six humeral head components, yielding six articulations with zero, one, two, three, four, or five millimeters of mismatch between the radii of curvature of the glenoid and humeral head components. The actual curvature of each humeral head and glenoid component was measured with a custom-made profilometer on an MTS mechanical testing machine (Minneapolis, Minnesota)16.
With the standard technique for clinical prosthetic implantation, the subscapularis tendon is incised and repaired and the anterior-inferior capsular ligament complex often is excised8,20. Healing of the tendon and the ligament is necessary for a functional structure after the operation. Because we wanted to minimize operative alterations of the length of the normal capsular ligament tissue during implantation in this cadaver model, an alternative procedure was developed to minimize soft-tissue damage.
An incision was made through the capsular interval between the subscapularis and supraspinatus tendons (the rotator-interval region of the capsule). The entire humerus, including the head, was split longitudinally through this incision to the distal part of the humerus with use of a power sagittal saw. The shaft was opened like a book so that the inside of the joint could be viewed, and a humeral osteotomy template was used to score both halves of the head. The sagittal saw was used to resect each half of the humeral head by starting at the score and proceeding to the edge of the articular surface in both halves. This ensured that the plane of resection approximated the anatomical retroversion of the humeral head. With removal of the head, the glenoid was easily accessible, with little operative alteration of the joint capsule. The glenoid surface was reamed so that its curvature matched that of the back of the glenoid component, and a keeled glenoid component was fixed in place with polymethylmethacrylate (Zimmer, Warsaw, Indiana) and standard cementing techniques8,20.
Marrow was removed from the medullary canal of the humerus, and polymethylmethacrylate was inserted into both halves of the shaft. The humeral stem was positioned so that the lateral fin came to rest approximately three to five millimeters posterior to the bicipital groove. The two halves of the shaft were brought together and held in place with plastic cable ties, producing a rigid and anatomical configuration.
After this procedure, the joint capsule was intact, except for a longitudinal incision in the rotator interval. A humeral head component was inserted into the joint through this incision, and the interval was anatomically closed with sutures. Once testing with a particular humeral head had been completed, the rotator interval was reopened by removing the sutures. The head was extracted, another was inserted, and the interval was again closed with sutures. This procedure was repeated for all six of the humeral head components for each specimen. Before insertion, each head was coated with K-Y lubricating jelly (Johnson and Johnson, New Brunswick, New Jersey) to reduce friction between the metal and plastic components.
Motion Analysis
Coordinate System
Joint motions were monitored with a six-degrees-of-freedom magnetic tracking device (3Space Fastrak System; Polhemus, Colchester, Vermont)1,19. The transmitter of the device was fixed to the scapula, and a scapular axis was defined on the basis of the plane of the scapula (two points on the medial border and one at the center of the glenoid curvature)16. The receiver of the device was fixed to the humerus, and a humeral axis was defined so that it was aligned with the scapular axis when the humerus was in its neutral position—that is, with the center of the humeral head coinciding with the center of the glenoid, the extremity at the side, and the humeral condyles parallel to the plane of the scapula. Translations were recorded along the superior-inferior, medial-lateral, and anterior-posterior axes. Euler angles were used to represent three sequence-dependent rotations: the plane of elevation, the degree of elevation, and internal-external rotation, as described by An et al.2.
This procedure was used to collect the data for both natural and reconstructed joints. Arthroplasty did not alter the alignment of the medial border of the scapula or the neutral orientation of the humerus; it did, however, alter the centers of the glenoid and the humeral head. Therefore, it was necessary to define different coordinate systems for each reconstructed joint. This was done once the entire experimental protocol had been completed, as it required complete excision of the joint capsule. Thus, all data were saved in an uncalibrated format and subsequently were calibrated with the appropriate axis system.
Center of the Glenoid Component
Once the entire experimental protocol had been completed, the capsule was excised. In order to find the center of the implanted glenoid component, the conforming humeral head (the component with a twenty-five-millimeter radius of curvature) was placed on the humeral stem, and the humerus was rotated through a central range of motion (one in which the ligaments were not taut), with the head centered in the glenoid component. The point that moved the least in the scapular reference frame was defined as the center of the glenoid component. As the same glenoid component was used for all of the experiments performed after the arthroplasties, this center was used for the coordinate-axis system for all of the implants.
Center of the Humeral Head Component
To determine the center of the humeral head component, it was necessary to articulate the head with a conforming glenoid component. Because only one of the humeral heads had a radius of curvature (twenty-five millimeters) that conformed with that of the glenoid component, additional glenoid components were machined to fit snugly over the implanted component. Each of these so-called calibration glenoid components matched the radius of curvature of one of the six heads. In turn, each head was rotated passively on its matching calibration glenoid component, and the point that moved the least in the humeral reference frame was defined as the center of that humeral head component.
Subtraction of Systematic Errors
As the magnetic tracking device is susceptible to error when it is in the vicinity of a metal object10,12, we attempted to increase its accuracy by subtracting systematic errors. This involved articulating the humeral head under conditions of no translation. Therefore, the passive experimental protocol was repeated for each humeral head by articulating it with its corresponding calibration glenoid component after the entire capsule had been excised. The translations that were recorded during these motions then were subtracted from the translations that were recorded during the actual experiment.
Errors in Measurement
Although the accuracy of the magnetic tracking device for these experiments in the absence of metal was found to be approximately 0.4 millimeter at the 95 per cent confidence level16, the effect of metal in the environment was assessed in the present study. After the capsule had been excised, the passive experimental protocol was repeated with the conforming (twenty-five-millimeter) humeral head on the original, implanted glenoid component. Because there should be no translation during this so-called zero translation motion, translations that were recorded after subtraction of systematic errors were considered an indication of random error.
Experimental Protocol
The glenohumeral joint was positioned in maximum internal rotation and then externally rotated in 10-degree increments until maximum external rotation was achieved. This protocol was chosen because both internal and external rotation are positions in which the capsular ligaments are presumed to be tight. A previous study of the motions of the natural glenohumeral joint demonstrated that the patterns of translations during active and passive internal-to-external rotation motions were similar for various planes and elevations16. Therefore, only one position was used for testing in the present study: 90 degrees of total elevation (60 degrees of glenohumeral elevation and 30 degrees of scapulothoracic elevation) in the plane of the scapula.
At each increment of external rotation, the elevation and the plane were maintained within 5 degrees of the desired values, and the rotation was maintained within 1 degree of the target value. Experiments were conducted first on the natural joint, after which the arthroplasty was performed and testing was repeated with each of the six humeral head components in place, in a random order.
Active and passive motions were performed in a random order on the natural joint and each implant. Active motions were achieved by applying forces to the sites of the insertions of the muscles with use of the hand cranks and spring scales. A 2.3-kilogram mass was hung on the distal end of the humerus to approximate the weight of the upper extremity. Maximum rotation during active motions was set with a sixty-newton force on either the infraspinatus-teres minor complex or the subscapularis.
For passive motions, the medially applied centering load was achieved by the application of a ten-newton force to each of the simulated rotator-cuff muscles with use of hanging weights and pulleys. Maximum rotation during passive motions was set by the application of torque through a torque wrench attached to the end of the humeral shaft. In order to approximately reproduce the torque that had been achieved actively (a sixty-newton force multiplied by a twenty-five-millimeter moment arm), a maximum torque of 1.5 newton-meters was set for both internal and external rotation.
On completion of the experimental protocol, the anterior and posterior bands of the inferior glenohumeral ligament were identified by palpation. By following along the length of these bands, we were able to identify their sites of osseous origin and insertion, and these were digitized with the stylus of the magnetic tracking device. After the location of the origin in the scapular axis and the insertion in the humeral axis had been determined, the distance between these two points was calculated from the data collected with the tracking device during experimentation. This method allowed us to determine the lengths of the bands for each experimental position13.
The range of motion for a given experiment was defined as the angle between the points of maximum internal and external rotations. The net anterior-posterior and superior-inferior translations were defined as the maximum translations over the range of motion for that experiment.
Statistical Analysis
SYSTAT (Version 5.2 for the Macintosh computer; Systat, Evanston, Illinois) was used for statistical analysis. A repeated-measures analysis of variance was performed for each dependent variable with two within-subject factors: model (active and passive) and experiment (the natural joint and the six implants). A Bonferroni correction factor was used to account for the fact that a total of five analysis-of-variance tests were performed (range of motion, anterior-posterior translation, superior-inferior translation, and the lengths of the anterior and posterior bands of the inferior glenohumeral ligament). A priori comparisons (or contrasts) between group means were planned before the experimentation. Unlike post hoc tests, which essentially make all possible comparisons, contrasts allow for fewer comparisons among means22. To determine the effects of arthroplasty, contrasts were made between the value for the natural joints and the group mean for the reconstructed joints. To determine the effects of alteration of the articular conformity, polynomial contrasts were used to test for linear trends among the implants. Because the contrasts used in the present study were orthogonal, or independent, a Bonferroni correction factor was not used18,22. The acceptable rate for a type-I error was chosen as 5 per cent (p = 0.05) for all tests.
Accuracy of Measurements
Translation
The net anterior-posterior and superior-inferior translations for the so-called zero translation motion (articulation of the humeral head component with a conforming glenoid component after capsular excision) were calculated for each specimen. The range of motion used was set at the maximum active and passive range of motion for that specimen. After subtraction of systematic errors, translation errors (at the 95 per cent confidence level) were 0.5 millimeter for passive motions and 0.3 millimeter for active motions.
Radius of Curvature
Measurements of articular geometry were made for all of the glenoid and humeral head components used in the present study. The mean difference between the measured and nominal values for the radius of curvature for both components was less than 2 per cent. Therefore, the nominal values were used in all of the analyses of the data.
Arthroplasty Procedure
Because all of the humeral components had a neck of the same length, each reconstructed joint was moved laterally by approximately the same amount (mean, seven millimeters; range, three to twelve millimeters) at 60 degrees of glenohumeral elevation in the scapular plane and the upper extremity in neutral rotation. Although the length of the neck was kept constant, the width of the neck and the centers of the articular surfaces varied as a result of changes in the radii of curvature. The humeral components were found to have a mean retroversion angle of 26 degrees (range, 10 to 42 degrees) and a mean neck-shaft angle of 41 degrees (range, 34 to 48 degrees).
Comparison of Active and Passive Models
The mean range of motion of the glenohumeral joint was found to be greater (p = 0.0005) during passive motions than during active motions (134 compared with 85 degrees). In fact, the passive range of motion exceeded the active range of motion in all experiments on every specimen. Along the anterior-posterior axis, the mean translations were also greater (p = 0.05) in the passive model (6.3 millimeters) than in the active model (1.1 millimeters). As with the data for range of motion, the passive translations exceeded the active translations in all experiments on every specimen. In both the natural and the reconstructed joints, these larger passive anterior-posterior translations occurred at the extremes of internal and external rotation—regions of motion that are not achievable actively. Anterior translation occurred at the extreme of internal rotation, and posterior translation occurred at the extreme of external rotation (Fig. 1). Although the mean superior-inferior translations also were greater in the passive model (2.5 millimeters) than in the active model (0.9 millimeter), these results were not found to be significantly different (p = 0.075), with the numbers available. In addition, the patterns of superior-inferior translation at the extremes of motion were not as consistent as the anterior-posterior translations.
Active Model
Comparison of Natural and Reconstructed Joints
In general, the data for active translation and the range of motion were similar for the natural and reconstructed joints. With the numbers available, we could detect no significant difference (p > 0.5) between the mean active range of motion before the arthroplasty (83 degrees) and that after the arthroplasty (85 degrees) (Fig. 2). In addition, no significant difference (p > 0.5) was detected between the mean superior-inferior translations before (1.1 millimeters) and after (0.8 millimeter) the arthroplasty (Figs. 3) or between the mean anterior-posterior translations before (1.5 millimeters) and after (1.0 millimeter) the arthroplasty (Fig. 4).
Effects of Articular Conformity
Changes in the active range of motion could not be accounted for by increases in the radial mismatch between the components of the prosthetically reconstructed joints (p = 0.2) (Fig. 2). Significant linear increases, however, were observed in both superior-inferior (p = 0.01) and anterior-posterior (p = 0.0015) translations (Fig. 3 and 4; the relative increases were 175 per cent (from 0.4 to 1.1 millimeters) and 467 per cent (from 0.3 to 1.7 millimeters), respectively. Along the anterior-posterior axis, the progressive increase in active translation with increasing radial mismatch was due to the tendency of the humeral head to progressively translate anteriorly with increased external rotation over the full active range of motion (Fig. 1). This pattern was observed in all specimens.
Passive Model
Comparison of Natural and Reconstructed Joints
There were variations between the natural and reconstructed joints during passive motions. With the numbers available, we could detect no significant difference (p > 0.5) between the mean range of motion of the natural joints and that of the reconstructed joints in the passive model (142 compared with 133 degrees) (Fig. 2). The mean passive superior-inferior translation was reduced, significantly (p = 0.025), from 4.3 millimeters before the arthroplasty to 2.2 millimeters after it (Fig. 3). The mean passive anterior-posterior translation was also reduced, from 9.3 millimeters before the arthroplasty to 5.8 millimeters after it, but this reduction was not found to be significant (p = 0.3) (Fig. 4).
Effects of Articular Conformity
A linear increase (p = 0.0018) in the range of motion was found with increasing radial mismatch in the passive model (Fig. 2). This increase (from 120 to 146 degrees) was more than twice that seen in the active model. With the numbers available, we could detect no significant change in either superior-inferior (p > 0.5) or anterior-posterior (p = 0.2) translations with increasing radial mismatch in this model (Figs. 3 and 4). The patterns of both superior-inferior and anterior-posterior translations varied among the specimens, with the translation increasing in some specimens and decreasing in others as articular conformity was decreased.
Ligamentous Constraints
The length of the anterior and posterior bands of the inferior glenohumeral ligament was calculated as the shortest distance between the sites of origin and insertion. The anterior band lengthened with external rotation and the posterior band lengthened with internal rotation in all specimens.
The length of the anterior band at maximum external rotation was significantly greater (p = 0.045) during passive motions than during active motions (Fig. 5). There was a slight but significant decrease (less than 4 per cent) in the length of this band with increasing radial mismatch during both active (p = 0.04) and passive (p = 0.025) motions.
Similarly, the length of the posterior band at maximum internal rotation was significantly greater (p = 0.035) during passive motions than during active motions (Fig. 5). However, with the numbers available, we could not find a significant effect of radial mismatch on the length of the posterior band during either active or passive motions (p > 0.5 for both comparisons).
The relationship between radial mismatch—that is, articular non-conformity—and glenohumeral translation during active positioning of the joint after prosthetic arthroplasty was found to be similar to the relationship demonstrated in the natural, unreconstructed joint16,17. Articular conformity appears to be the predominant factor controlling translations during these motions. Relative increases in active superior-inferior and anterior-posterior translations of 175 and 467 per cent, respectively, were observed when the radial mismatch of the glenohumeral components increased from zero to five millimeters. Although decreasing articular conformity led to increased active translations, the mean active translation was less than two millimeters with the least conforming components. On the basis of theoretical calculations15, it appears that these translations are less than what is required to achieve contact between the humeral head and the glenoid rim. The muscle forces required for active positioning of the joint, therefore, acted to limit glenohumeral translations to an amount that prevented loading of the glenoid rim.
Glenohumeral translations after prosthetic arthroplasty were larger during passive positioning of the joint than during active positioning. Also, translations varied more among specimens during passive positioning than during active positioning, as evidenced by the larger standard errors of the mean for both superior-inferior and anterior-posterior translations (Figs. 3 and 4). Active translations tended to occur over the entire range of motion, during which the head was in contact with the glenoid surface, whereas passive translations were observed primarily during ranges of motion that were not achievable during active range of motion. Consequently, no significant relationship was found between passive translations and articular conformity, with the numbers available. However, increased passive translations were associated with increased length of the inferior glenohumeral ligament, with the increased translation occurring in the direction opposite to that of the elongating ligament. For example, internal rotation during passive motions produced lengthening of the posterior band of the inferior glenohumeral ligament, which was associated with an increase in anterior translations.
The patterns of translation were similar for the natural and reconstructed joints, with the only significant difference being a decrease in the net passive superior-inferior translation after arthroplasty. There were large specimen-to-specimen variations in the active translations of the natural joint. This variability probably was due to differences in glenohumeral conformity among the specimens, as was documented in a previous study of natural joints16. Because of this variability and the small number of specimens used in the present study, it is difficult to predict the amount of conformity of the glenohumeral components that is needed to reproduce the translations of a natural joint most accurately. For example, to reproduce the patterns of anterior-posterior translation of the natural joints on an individual basis, the amount of articular congruence would range from perfectly conforming to a radial mismatch of more than five millimeters. On the average, however, active translations of the natural joint were most similar to those of the reconstructed joints that had a radial mismatch of approximately four millimeters (Figs. 3 and 4).
During simulated active motions, the articular surface of the humeral head presumably was in contact with the articular surface of the glenoid; this explains the influence of articular conformity on translations. To accommodate the larger passive translations, however, either the humeral head was not in contact with the glenoid or its edge or beveled portion was in contact with the glenoid, as was discussed by Ballmer et al. In these cases, the capsular structures had a more important role in the control of translations, as the anterior and posterior bands of the inferior glenohumeral ligament were longer at maximum rotation during passive motions than during active motions.
Few investigators have addressed the effects of articular conformity on forces and translations after total shoulder arthroplasty. Severt et al. studied specimens that had been excised from cadavera and found that, when articular conformity and constraint were reduced, joint subluxation forces also were reduced. Harryman et al. found that articular conformity did not have a significant influence on glenohumeral translations in a passive cadaveric model11. However, those authors did not examine the effects of maintaining a compressive load on the joint. The results of the present study indicate that articular conformity plays a greater role in the control of translations when muscle forces are present.
Although the technique for arthroplasty that was used in the present study is clinically unconventional, it maintains the integrity of the capsular ligaments; therefore, it was appropriate for this experimental model. The mean 26-degree retroversion angle for the humeral shaft in the present study is only slightly smaller than the 30 to 35-degree angle that has been typically recommended for this procedure20. Similarly, the mean neck-shaft angle of 41 degrees in the present study is very close to the 45-degree angle that has been reported for the natural joint14. The movement of the humerus laterally due to reconstruction of the joint also occurs in the clinical situation and is primarily a result of the thickness of the glenoid component and the position of the humeral stem.
In the clinical environment, prosthetic replacement often is performed in a shoulder with capsular contracture and rotator-cuff disease. Operative soft-tissue balancing and rotator-cuff repair are attempts to restore these tissues to a more natural condition. Clearly, however, re-establishing normal soft-tissue balance and rotator-cuff function is a goal that is rarely, if ever, achieved. Because each of the cadaveric specimens in the present study had intact rotator-cuff muscles and normal capsular tissues, it was desirable to avoid alteration or damage of these tissues so that different patterns of motion could be attributed to changes in the characteristics of the implant rather than to deficient soft tissue. This was accomplished with the modified technique for arthroplasty.
The specimens used in the present study showed no signs of injury of the rotator cuff or capsular contracture. This often is not the case for patients who have a shoulder arthroplasty. Depending on the type of arthritis, there may be substantial capsular contracture or rotator-cuff injury. Although these conditions often are improved through capsular release and repair of the cuff with the standard technique for total shoulder arthroplasty, it is unlikely that the soft tissues are returned to their normal condition. Previous studies of cadavera have demonstrated that capsular contracture (or capsular imbrication) increases translations of the humeral head during both active and passive positioning of the joint4,10. A similar study also showed increased translations due to rotator-cuff injury in an active model23. Because our study was designed to characterize the relationship between glenohumeral motions and conformity of the components, independent of problems with the capsule and the rotator cuff, we believed that simulating these conditions would only weaken the study.
The limitations of the current study include those that have been discussed previously, such as the somewhat arbitrary selection of muscle forces, the difficulty in interpreting data on the length of the ligament, and the venting of the joint capsule16. With regard to the accuracy of measuring translation, the magnetic tracking device (the 3Space Fastrak), because of its pattern of systematic errors, was found to be as accurate in measuring translations of a reconstructed joint as it is in measuring translations of a natural joint, even in the presence of metal. Although clinical knowledge and electromyography were used to determine the forces in the present study, other combinations of muscle forces may have produced the same motions. The lengths of the ligament that are reported in the present study are reflections of the straight-line distance between the sites of origin and insertion and do not account for wrapping of the ligament around the articular surface of the humeral head. Finally, it was not possible to perform the arthroplasty without venting of the joint.
In conclusion, articular conformity was found to play a vital role in the control of glenohumeral translations during active positioning of the joint. This is an important starting point for improving the understanding of the biomechanics of shoulder arthroplasty systems. The consequences of these translations, however, still need to be investigated.