In the absence of any established animal model suitable for in vivo testing of different techniques for rotator cuff repair, the shoulder of the Alpine sheep was selected for this experiment because of the similarity of its infraspinatus tendon to the human supraspinatus tendon2,4. The left infraspinatus tendons of forty-seven adult female Alpine sheep were used. The mean weight (and standard deviation) of the sheep was 60 ± 7 kilograms (range, forty-four to seventy-five kilograms), and the mean age was 5 ± 2 years (range, two to nine years).
Test Groups
Three test groups were compared (Table I). Group A had repair of the infraspinatus tendon with three simple stitches and the conventional double transosseous bone-fixation technique. Group B had repair with two modified Mason-Allen stitches and the conventional double transosseous bone-fixation technique. Group C had repair with two modified Mason-Allen stitches and the cortical-bone-augmentation double transosseous bone-fixation technique. Use of the conventional double transosseous bone-fixation technique in both group A and group B allowed comparison of the two different tendon-suturing techniques, whereas use of the modified Mason-Allen stitch in both group B and group C allowed comparison of the two different bone-fixation techniques. The Investigational Review Board at our institution believed that a fourth test group consisting of repair with simple stitches and the cortical-bone-augmentation technique was not warranted because the first three groups would allow identification of whether the modified Mason-Allen stitch was better than the simple stitch and whether the augmented transosseous bone-fixation technique was superior to the nonaugmented transosseous bone-fixation technique.
All of the sheep were suspended in a sling postoperatively and protected from full weight-bearing. The involved limbs were started moving immediately postoperatively through an increasing range of motion. The sheep were observed to be calm postoperatively and did not visibly suffer. No analgesics or nonsteroidal anti-inflammatory drugs were used.
Development of the Sheep Model
Initially, we planned to study the delayed repair of tears of the infraspinatus tendon four to six weeks after transection of the tendon, which would have simulated features of a chronic rotator cuff tear, including retraction, atrophy, and loss of compliance of the infraspinatus muscle as well as osteoporosis of the greater tuberosity of the humerus. In addition, to simulate the condition in humans (in whom some severely retracted rotator cuff tendons can be repaired only under tension), we planned to shorten the infraspinatus tendons before the repair to create tension at the site of tendon-bone fixation. However, a variety of problems were encountered in these pilot studies. The initial protocol was therefore modified several times, and the delayed-repair model was abandoned in favor of an immediate-repair model, with the infraspinatus tendon transected and repaired during the same procedure. Nonetheless, the observations made during the course of these pilot studies have strong clinical relevance and are described in the Results section.
Characteristics of Unsutured Tears of the Infraspinatus Tendon in Sheep
To create a model of delayed repair, a tear of the infraspinatus tendon was made in the left shoulder of two sheep by releasing the tendon at its insertion. The sheep were kept partially immobilized in a hanging device for one week postoperatively to avoid excessive motion of the affected limb. The hanging device consisted of belts around the abdomen and thorax of the sheep. The belts were fixed to the ceiling of the stable with ropes. This hanging device allowed the sheep to stand and bear full weight but prevented them from walking or lying down. When sleeping, the sheep were suspended in these devices. After four to six weeks, the shoulders were removed and dissected. Scar tissue covered the tendon-bone junction. The released infraspinatus tendons were retracted by two to three centimeters, and the scar tissue filled the defects and embedded the tendon stumps. The infraspinatus musculotendinous units were completely removed from the shoulders and weighed with a balance. The volume of the units was determined by measuring the displaced volume of water in a container. Compared with the musculotendinous unit on the contralateral side, the involved musculotendinous units of the two sheep had lost 17 and 18 percent of their weight and 12 and 15 percent of their volume.
Simulation of Osteoporosis by Decortication of the Greater Tuberosity
In each of the two sheep shoulders that had the four to six-week-old iatrogenic tear of the infraspinatus, the humeral head was hard and of good bone quality. Bone-density measurements made on a computed tomography scan revealed a quantitative loss of bone of only 10 and 17 percent at the greater tuberosity compared with the contralateral side. This loss is less than that observed in the humeral heads of patients who have long-standing disease of the rotator cuff4. To simulate osteoporosis in humans, the cortical bone of the greater tuberosity was removed with a chisel until cancellous bone was exposed. The area of weakened bone was four centimeters by one centimeter in diameter and was located one centimeter lateral to the site of the original insertion of the infraspinatus tendon (Fig. 3). In vitro pullout tests of double transosseous sutures were performed on eight sheep humeral heads. The testing and loading conditions were identical to those in a previously reported in vitro study in which the same tests were performed on six osteoporotic humeral heads from human shoulders with a massive rotator cuff tear4. In our study, two Ethibond number-3 sutures pulled through the weakened sheep bones at a load of 165 ± 56 newtons, whereas they pulled through the six osteoporotic human humeral heads in the previous study4 at a load of 146 ± 41 newtons. With the numbers available, no significant differences were found between these two values with the Mann-Whitney U test (p = 0.3).
Pilot Studies of Delayed Repair of Tears of the Infraspinatus Tendon
In the pilot studies of the delayed repair of tears of the infraspinatus tendon, the tendon was first transected at the greater tuberosity and then repaired after a delay of four to six weeks.
All procedures were performed with the sheep under general anesthesia (thiopental, halothane, and nitrous oxide).
A lateral approach was used to retract the deltoid muscle medially. This procedure widely exposed the infraspinatus tendon, which was released at its insertion. The wound was then closed in layers. Postoperatively, the sheep were kept in the hanging device until the second operation for the repair of the tendon.
For the transosseous repair of the tendon after four to six weeks, the same approach was used. The infraspinatus tendon was mobilized by releasing the peritendinous scar tissue. To simulate the intraoperative experience in humans (in whom severely retracted tendons of the rotator cuff can often be repaired only under tension), we shortened the stump of the infraspinatus tendon until an intraoperative tension of approximately thirty newtons was obtained. This level of tension was chosen arbitrarily but corresponds to a level that we have often measured during rotator cuff repairs in humans. The amount of shortening was five to ten millimeters to obtain thirty newtons of tension, which was measured with the extremity in neutral position with a spring-balance (PESOLA; Präzisionswaagen, Baar, Switzerland). Two modified Mason-Allen stitches or three simple stitches with Ethibond number-3 sutures were then placed into the tendon ten millimeters proximal to the tendon end (Fig. 3). A five-millimeter-deep bone trough was created at the site of the original insertion of the infraspinatus tendon. The bone of the greater tuberosity was weakened with decortication as described earlier, and the sutures were pulled through three two-millimeter drill-holes. Either the nonaugmented or the augmented bone-fixation technique was used. For the latter technique, the cortical bone was augmented with two two-millimeter-thick, fifteen-millimeter-long, and ten-millimeter-wide absorbable poly(L/D-lactide) plates (G. HUG, Freiburg-Umkirch, Germany).
Two metallic vascular clips were then placed into the tendon stumps and into the greater tuberosity so that the amount of gap formation between the tendon and the proximal end of the humeral head could be monitored radiographically. The wound was closed in layers. The sheep were kept in the hanging device and protected from full weight-bearing for the first postoperative week. Thereafter, no additional protection was used.
The distance between the clips was measured intraoperatively with a ruler and on radiographs immediately postoperatively; after one day; after two, four, six, eight, and ten weeks; and at the end of the experiment.
The sheep were killed after six weeks, three months, or six months or when a gap of fifteen millimeters or more between the two clips was discovered on the postoperative radiographs.
The shoulders were removed, and the humeral heads and the tendons were macroscopically analyzed for integrity of the repair. Eleven involved shoulders were used for mechanical pullout tests, and eight contralateral shoulders were used as a control group. For the mechanical tests, the humeral heads were freed of all soft tissues except the infraspinatus tendon and were embedded in methylmethacrylate. The infraspinatus tendon was grasped with a specially constructed clamp. The pullout tests were performed with a universal testing machine (model 4302; Instron, High Wycombe, England). No preload was used for the mechanical tests. The extension rate was five millimeters per minute. During the testing, the specimens were kept moist with saline solution.
Twenty-four specimens were analyzed histologically by one of us (R. W. N.) with the help of two board-certified pathologists. Immediately after the sheep were killed, the axillary artery was injected with a 4 percent dispersion of India ink in Ringer's lactate. The specimens were then submerged into a 4 percent formalin solution, decalcified with 15 percent formic acid, and prepared step by step with water, alcohol, and xylol. They were then embedded in methylmethacrylate and cut longitudinally in the direction of the fibers of the infraspinatus tendon vertical to its surface, in slices of six micrometers, at intervals of five millimeters.
Light microscopy with polarized and nonpolarized light was used to histologically evaluate the specimens after staining with Giemsa and van Gieson stains.
Statistical Analysis
After a discussion with our statistical consultant, we decided not to use statistical analysis to evaluate the data because strictly comparable test groups were too small after the repair methods had been modified several times.
Four sheep were excluded: three had a deep wound infection, and one died of aspiration during the operation. No antibiotics were used during the experiments.
For reasons that will be described, we observed an inordinately high rate of failure in the first eight sheep treated with delayed repair in the pilot study (Table I, series I). The experiment was therefore interrupted. After analysis of the results in these sheep, the methods were modified as will be described. The next step in the experiment involved seven sheep (Table I, series II) and also led to a high rate of failure, for reasons that will be described, so the methods had to be further modified. We made a total of four modifications to the methods before we recognized that protection from full weight-bearing was absolutely essential and that it was necessary to develop an efficient method of prevention of full weight-bearing. There were thus five series: one representing the original method and four, a modification of the method. The observations made during these pilot studies are of strong clinical relevance and are therefore described.
Observations in Series I (Table I)
In series I (delayed repair), seven of the eight repairs, involving each of the three repair techniques, failed. Histological examination revealed that the stitches in the delayed repair had inadvertently been placed into peritendinous scar tissue and not into the retracted original tendon tissue. At the time of the delayed repair, this scar tissue had been indistinguishable from normal tendon. Subsequent in vitro pullout tests of modified Mason-Allen stitches placed into the peritendinous scar tissue of two specimens that had a failure of the repair yielded low pullout strength (sixty-four and seventy-eight newtons). This series of experiments was excluded from further analysis, and the protocol was modified as will be described.
Modification of Methods and Observations in Series II Through V (Table I)
In series II (seven sheep; delayed repair), the sutures were placed twenty to twenty-five millimeters proximal to the distal end of the tendon to ensure that the stitches were in the original tendon tissue. In these repairs, the tendon-scar stumps were sutured into the bone trough. Resection of all scar tissue at the distal end of the tendon would have caused tensions that were too high to allow repair to the bone trough. The radiographic follow-up of these sheep indicated failure of the repair in five: two in group A, one in group B, and two in group C. After the sheep were killed, histological analysis did not allow the differentiation of old scar from new scar. With this very high rate of failure, we decided to abandon the delayed repair in favor of an immediate repair of a released tendon.
In series III (nine sheep), we used a one-stage procedure with release of the infraspinatus tendon at its insertion, partial resection of the distal end of the tendon, and immediate repair under thirty newtons of preload. The musculotendinous unit was more compliant and was not retracted or atrophic. Also, direct repair of the tendon end was easily performed. The radiographic follow-up, however, revealed failure of the repair in eight of the nine sheep in this series despite the good compliance of the muscles and the ease of repair.
In series IV (nine sheep), we decided to further reduce the postoperative load. No tissue from the distal end of the tendon was resected, and the repair was performed under no tension. Seven of the nine repairs in this series failed.
At this stage of the experiment, it became clear that, regardless of technique, none of our tested repairs would withstand the loads imposed under unprotected experimental conditions. Protection from weight-bearing was essential. The hanging device still allowed full weight-bearing, and full suspension or immobilization of the limb (including the upper body) in a cast was not tolerated by the sheep. Thus, in series V (ten sheep), a hard ten-centimeter-diameter rubber ball was fixed under the hoof of the involved limb with methylmethacrylate. With this device, the sheep did not bear full weight on the limb but touched the ground, avoiding full weight-bearing and reducing rotational forces between the limb and the ground. The rubber ball was removed after five weeks.
Integrity of Repairs
In series II, III, and IV (without postoperative prevention of weight-bearing), all of the repairs in group A (simple stitch and nonaugmented bone fixation) failed with a gap of more than fifteen millimeters. The cause was pullout of all three simple stitches from the tendon in four of the eight shoulders. In two shoulders, all of the transosseous sutures were found to be pulled into the humeral head. In the remaining two shoulders, a combined failure of the tendon suture and bone fixation was observed. The suture material never broke in this group.
In the group-B sheep (modified Mason-Allen stitch and nonaugmented bone fixation) of series II, III, and IV, two of the eight repairs healed (Fig. 4). All of the modified Mason-Allen stitches were found to be intact. In two limbs both sutures ruptured, and in four limbs one suture ruptured and the other pulled through the bone.
In the group-C sheep (modified Mason-Allen stitch and cortical-bone augmentation) of series II, III, and IV, three of the nine repairs healed. The modified Mason-Allen stitches slipped out of the tendon in two sheep. In four sheep, the suture material ruptured. The bone fixations were always intact.
Thus, a comparison of technique A (simple stitch) and B (modified Mason-Allen stitch), both with the nonaugmented double transosseous bone fixation, revealed that the simple stitches pulled out of the tendon in six of the eight shoulders, whereas all of the modified Mason-Allen stitches of group B were intact. Comparison of techniques B (nonaugmented double transosseous bone fixation) and C (augmented double transosseous bone fixation), which both used the modified Mason-Allen stitch, revealed failure of the nonaugmented bone fixation in four shoulders, whereas the augmented fixation was always intact.
Overall, without postoperative prevention of weight-bearing, all of the repairs with use of both simple stitches and nonaugmented double transosseous bone fixation failed. The modified Mason-Allen stitches were found to be intact in fifteen of seventeen shoulders. In the two repairs that failed, the modified Mason-Allen stitches slipped out of the tendon. The bone fixation was always intact when the augmentation had been used.
Series V (with prevention of full weight-bearing) consisted only of repairs with use of the modified Mason-Allen stitch combined with the poly(L/D-lactide)-plate bone augmentation. Eight of the ten repairs healed. In the two repairs that failed, the modified Mason-Allen stitches slipped out of the tendon.
Of the twenty-two failures in series II through V, one was identified on a radiograph immediately postoperatively; fourteen, after one week; three, after two weeks; three, after three weeks; and one, after six weeks. Thus, almost all of the failures occurred within the first three weeks after repair.
Macroscopic evaluation of the failed repairs showed that the gaps between the stump of the infraspinatus tendon and the bone trough were consistently filled with scar tissue that bridged the defect. We termed this type of rupture a failure in continuity.
Mechanical Testing of Repairs with the Modified Mason-Allen Stitch and Bone Augmentation
The eight normal infraspinatus tendons in the control group withstood a mean load (and standard deviation) of 2500 ± 286 newtons (range, 2221 to 3141 newtons). Failure at maximum load occurred by rupture of the tendon tissue at the site of the clamp, which suggests that fixation with the clamp may have weakened the tendon and that the actual strength of a normal tendon might be even higher.
Mechanical testing was performed on seven intact repairs, all of which were from group C (repair with the modified Mason-Allen suturing technique and bone augmentation). One was from series II; one, from series IV; and five, from series V (protection from weight-bearing) (Fig. 5 and Table I). In these seven successful repairs, the ultimate failure strength averaged 755 newtons (30 percent of that of normal tendon) after six weeks, 1291 newtons (52 percent) after three months, and 2030 ± 99 newtons (81 percent) after six months.
Mechanical testing was also performed on four unsuccessful repairs that had healed in continuity, with scar bridging the gap between the retracted end of the tendon stump and the bone trough (Fig. 5). At six months, the mechanical strength of the scar that bridged the gap between the tendon stump and the bone was approximately the same as that of a successfully healed tendon-to-bone junction (Fig. 5).
Because of the small number of samples in each group in this preliminary study, statistical analyses of the differences in the mechanical strengths of the repairs in the different groups were not carried out.
Structural Changes of the Tendons
The two types of stitches caused similar histological changes in the tendon tissue. The simple stitch caused changes as far as two millimeters around the suture material and as far as the distal tendon stump distal to it. With the modified Mason-Allen stitch, the involved tendon area was larger: there were changes within the sutured tissue, as far as two millimeters around the suture material, and distal to the stitches at the distal end of the tendon.
At one week postoperatively, these changes consisted of a reduced number of vessels and fibrocytes. The cell nuclei were pyknotic or were not stained, suggesting cell death (Fig. 6). In addition, an increased number of rounded cells were found, particularly distal to the stitches. The architecture of collagen-fiber bundles was always preserved. Between the stitches, the tendon tissue appeared normal. Around the tendon stumps, granulation tissue was found that contained vessels, fibroblasts, and longitudinally oriented collagen fibers.
At two weeks postoperatively, the tendon seemed to be edematous with a distended space between the collagen-fiber bundles and in between an increased number of vessels and fibroblasts showing numerous mitoses.
At six weeks, the tendon stumps were still edematous. Fibroblasts and vessels were present in large numbers, and inflammatory cells were observed.
Three and six months after the operation, most repairs appeared to be in continuity either by direct contact between the tendon and the bone trough or by interposition of scar tissue. The tendon stumps were less edematous but still contained numerous fibroblasts and vessels. Within the region surrounded by sutures, the number of fibroblasts was still diminished but the collagen-fiber bundles were generally intact. The scar tissue had a more mature appearance and contained dense collagen-fiber bundles that were mainly parallel and longitudinal in orientation. In some six-month specimens, it was almost impossible to distinguish the scar tissue from the edematous distal tendon stump.
Tendon-Bone Junction
In the sheep, as in humans, the insertion of a normal tendon into bone was found to consist of four layers: tendon, noncalcified fibrocartilage, calcified fibrocartilage, and bone.
At six weeks postoperatively, scar tissue consisting of fibroblasts and parallel collagen-fiber bundles filled the gap between the end of the tendon and the bone trough. Multiple osteoblasts at the border of the trough formed new bone, embedding these collagen-fiber bundles. Some vessels were observed to extend from the scar tissue to the bone marrow. At three months, the bone troughs were found to be completely laid out with new bone. The scar tissue was dense and parallel in orientation. At six months (Fig. 7), the scar tissue was compact and consisted mainly of collagen-fiber bundles lying tightly side by side and resembling normal tendon tissue. The fibers were packed and fixed into the adjacent bone mass, which was covered by a layer of dense, noncalcified fibrocartilage similar to that found at a normal tendon-to-bone junction. Calcified fibrocartilage (as found in a normal tendon insertion) was not observed.
Bone Augmentation with Absorbable Poly(L/D-Lactide) Membranes
The poly(L/D-lactide) plates were covered by a thin capsule of nonreactive connective tissue with only a few inflammatory cells such as macrophages and giant cells. After six months, the plates measured approximately 100 micrometers. All of the plates were found to be intact, but some of the membranes had a small degree of bowing, which indicated load uptake.
More than half of the repairs of chronic, massive rotator cuff tears rupture again and fail3,5,6. As successful repairs that remain intact yield substantially better functional results than those that fail3,5,6,8, obtaining healing of the repair is of crucial importance if postoperative function, rather than relief of pain alone, is a major goal. Current clinical5,6,9 and experimental2,4 evidence suggests that the repair technique may play an important role in the prevention of failure.
In humans, the analysis of failures of rotator cuff repairs, except those necessitating revision, is limited to imaging studies such as arthrography, ultrasonography, and magnetic resonance imaging. Even in prospective clinical series, many variables such as the size, location, and age of the tear as well as the postoperative compliance of the patients render the analysis of failures difficult. Animal experiments are necessary to define conditions for successful repairs in humans.
To our knowledge, no one has previously reported a satisfactory experimental model with which to develop techniques that lead to successful repair. After thorough study of different possibilities and in vitro experimentation with sheep shoulders4, we developed an in vivo animal model of human rotator cuff tears. In our pilot series I and II (delayed repair), retraction and atrophy of the muscle were simulated in subacute rotator cuff tears but it was impossible to distinguish between the true tendon stump and the newly formed scar at the time of delayed repair. This problem severely compromised the experiment. Therefore, in series III, IV, and V, the tendon was detached and immediately repaired so that the experimental condition corresponded to an acute tear. Degeneration of the tendon and changes of mechanical properties of the musculotendinous unit, as encountered in long-standing tears in humans7, were not simulated in series III, IV, and V.
Osteoporosis of the greater tuberosity is often encountered in shoulders that have a long-standing rotator cuff tear. The intraoperative finding in patients that nonaugmented transosseous suture fixation to osteoporotic humeral heads can be mechanically weak was confirmed in a previous in vitro study4. Other authors have also identified the greater tuberosity as a potential location for transosseous suture failure1. They found that the strength of transosseous suture fixation can be increased by placing the sutures at sites more distal to the tip of the greater tuberosity or by tying the sutures over a wider bone bridge. The role of osteoporosis in the failure of rotator cuff repair in human shoulders has not been studied in vivo, to our knowledge. Considering that the causes of failure of rotator cuff repairs in humans are not known, we planned to simulate the features of rotator cuff tears that might be important factors in failures, including osteoporosis. Decortication caused weakening of the greater tuberosity of the sheep, resulting in in vitro holding strength for transosseous sutures similar to that of osteoporotic humeral heads. However, this model can only be considered an approximation of the human shoulder with a massive rotator cuff tear and consequent osteoporosis.
The potential for healing seems to be better for the sheep infraspinatus tendon than for the human supraspinatus tendon. The retracted infraspinatus tendons in failed repairs were always embedded in scar tissue that bridged the created defects, which resulted in lengthening of the tendon and shortening (retraction) of the muscle (failure in continuity). Although rotator cuff tears in humans are more likely to be characterized by a tissue defect, intraoperative observations by one of us (C. G.) led us to believe that this type of failure in continuity also exists in humans. This observation deserves further attention.
The need for protection of the shoulder after rotator cuff repair is extremely controversial. We expected that the repairs would be compatible with the limited activity of the sheep. Suspension in a sling allowed the sheep to bear full weight on all of the extremities but prevented them from walking. Inability to reach the ground was not tolerated by the sheep, and it was only in the last series of experiments that a solution was found to prevent the sheep from bearing full weight on the affected extremity. Mounting a rubber ball on the involved hoof apparently changed the proprioceptive feedback of the sheep so that they avoided weight-bearing.
The present study documented that the holding power of a modified Mason-Allen stitch was distinctly better than that of a simple stitch, as most of the simple stitches failed and only two of the modified Mason-Allen stitches failed. On histological examination, this new tendon-suturing technique was found to be biocompatible. Signs of tendon-tissue damage were identified but were limited to a small area of sutured tendon tissue. This tissue damage was found predominantly at one week postoperatively. Thereafter, the tendon had mostly recovered. The simple stitch was less strangulating. Nevertheless, the same local tendon damage was found in a small area surrounding the stitches. The increased area of tendon damage with the modified Mason-Allen stitch was temporary and had no apparent clinical effect.
The failure modes in this study also documented that fixation over augmented cortical bone was distinctly more reliable than that over nonaugmented cortical bone. The repairs with augmented fixation never failed at their knot or over the plate; the in vivo results with bone augmentation therefore confirmed those of the corresponding in vitro study4. The absorbable poly (L/D-lactide) membranes were well tolerated by the sheep, but advanced degradation of the membranes had not yet taken place; a longer duration of follow-up is necessary for analysis of tolerance of this type of implant. In clinical practice, a titanium plate is currently used and has not yet been associated with complications. A repair with a modified Mason-Allen stitch and cortical-bone augmentation transformed the suture into the weakest link in the chain.
Regardless of the repair technique, the suture material was not able to withstand the loads that were imposed on the repair if the sheep were not prevented from weight-bearing. However, without weight-bearing, the loads imposed on the repairs performed with the modified Mason-Allen stitch and cortical-bone augmentation could be managed successfully and the repairs healed.
Poor-quality tendon tissue may be another cause of failure of rotator cuff repairs. It was not possible to address this question in the present study. However, placing the stitches into peritendinous scar tissue resulted in disruption of the repair. The four to six-week-old scar tissue felt hard when the suture needle was passed through it, but mechanical tests showed that this immature scar tissue was weak. We therefore try never to put stitches into tendon-like tissue that is hard or stiff or appears myxomatous, and we usually debride such tissue, during rotator cuff repair in patients.
The model in the present study certainly cannot be accepted as fully simulating rotator cuff repair in humans. Nonetheless, the study documented that a modified tendon-suturing technique was associated with fewer failures than a simple stitch, without causing strangulation and tendon necrosis. It confirmed that nonaugmented fixation in the greater tuberosity could fail, whereas augmented bone fixation did not, and that the use of such a repair technique did not cause any apparent complications. Most of all, the study documented that, regardless of the technique, no repair was able to withstand the high loads imposed by weight-bearing. Given that the clinical results of successful repairs are superior to those of failed repairs3,5,6,8, the use of an optimum repair technique and adequate protection of the repair until healing appear warranted.
NOTE: The authors wish to acknowledge the enormous technical support of Roland Würgler, mechanical engineer, throughout the study. Thanks are also due to Jiangming Xu, M.D., Werner Wüst, M.D., and Hubert Laeng, M.D., for their contributions to the histological evaluation; to Iris Keller, Elena Rampoldi, Petra Romer, Katrin Kampf, and Margret Hostettler for the preparation of the histological specimens and operative assistance; and to Urban Lanker for the care of the sheep. The support of Prof. Berton Rahn, M.D., throughout this study is especially and gratefully acknowledged.