Latissimus dorsi tendon transfer for treatment of irreparable rotator cuff tears was first reported by Gerber et al. in 19881, to our knowledge. Subsequently, several authors described their experiences with this salvage procedure for the treatment of massive rotator cuff tears2-8 and considered it to be a reliable method. Predictors of failed outcomes were analyzed6, but only sparse information has been reported concerning the prevalence of rupture of the transferred tendon resulting in clinical failure. Gerber et al.4 did not observe any evidence of tendon insertion failure on clinical examination. They reported that it was extremely difficult to assess the integrity of the transferred latissimus tendon, although they did not use other investigational methods such as ultrasound or magnetic resonance imaging. Warner and Parsons8 reported a 36% rate of failure of healing, which they detected clinically without the use of imaging. Habermeyer et al.5 introduced a single-incision technique through a dorsal approach and reported unsatisfactory clinical results in 17% of their patients. Aoki et al.2 reported poor clinical results in 25% of their patients. Along with subscapularis deficiency and nonsynergistic action of the transferred muscle, rupture of the tendon at the greater tuberosity was assumed to be a possible reason for the poor results, and this was confirmed arthroscopically in one case. No imaging was used to evaluate the integrity of the transferred tendon. Miniaci and MacLeod7 observed an 18% rate of unsatisfactory outcomes in their series of latissimus dorsi tendon transfers performed as revision procedures.
We believe that Iannotti et al.6 were the first to perform magnetic resonance imaging in all of their patients to evaluate tendon healing, the degree of fatty infiltration, and atrophy of the muscle belly. They reported that 36% of their patients were dissatisfied with the result of the surgery, and magnetic resonance imaging demonstrated equivocal evidence of latissimus tendon healing to the greater tuberosity in 14%.
We reviewed our experience with latissimus dorsi tendon transfer performed with a modified technique of tendon harvesting that allows stable bone-to-bone fixation. We assessed the clinical and radiographic outcomes in two groups of patients. The first underwent a standard transfer of the latissimus dorsi tendon following sharp release from its humeral insertion. In the second group of patients, the latissimus dorsi tendon was detached from its insertion along with a small piece of bone to permit bone-to-bone transosseous fixation of the transfer.
Our hypothesis was that the modified technique of tendon harvesting with use of a small piece of bone could reduce the rupture rate as detected by magnetic resonance imaging and improve the final clinical outcome of latissimus dorsi tendon transfer.
Patient Data
Between 1997 and 2005, forty-seven patients with a chronic irreparable tear involving the supraspinatus and infraspinatus muscles of the rotator cuff were managed with a latissimus dorsi tendon transfer. A diagnosis of an irreparable rotator cuff tear depended not only on the tear size but also on the grade of fatty infiltration (3 or 4, as determined with the system of Goutallier et al.9). Three patients died of unrelated causes, and two patients were lost to follow-up. Thus, forty-two patients (89%) were included in this retrospective study. The indications for latissimus dorsi tendon transfer were (1) a clinically diagnosed massive rotator cuff tear involving the supraspinatus and infraspinatus tendons confirmed by findings on magnetic resonance imaging of grade-3 tendon retraction according to the classification system of Patte10 and grade-3 muscle atrophy according to the classification system of Thomazeau et al.11, or (2) a prior failed open repair of a posterosuperior rotator cuff tear. Twenty-four patients (57%) underwent tendon transfer for primary reconstruction, and eighteen patients (43%) underwent tendon transfer as a salvage reconstruction following failure of a prior rotator cuff repair. In the latter group, magnetic resonance imaging was performed before the index procedure to assess the size of the tear and the amount of tendon retraction, the grade of fatty infiltration, and atrophy of the muscle belly. In this series, a teres minor tear was not considered to be a contraindication to latissimus dorsi tendon transfer. The preoperative clinical examination revealed a positive hornblower sign, which may have been indicative of teres minor deficiency, in twenty-eight patients12. All patients reported continuous and chronic pain and weakness of external rotation in addition to weakness or complete loss of overhead elevation. The index procedure was not considered for patients with clinical evidence of subscapularis deficiency and a positive lift-off test13 because of the unfavorable outcomes reported by several authors under those circumstances2,4. A constrained inverse total shoulder prosthesis was recommended for those patients. Conservative treatment consisting of nonsteroidal anti-inflammatory drugs and physical therapy was tried for at least six months, even for those who had had prior surgery, and failed in all patients. All patients provided informed consent before undergoing the procedure.
Preoperative, intraoperative, and postoperative clinical and radiographic data were available from our shoulder database. The preoperative strength and range-of-motion measurements in both groups are detailed in Table I. The follow-up examination consisted of a detailed interview, physical examination, and radiographic examination. The patients were followed for a mean of forty-seven months (range, twenty-four to sixty-seven months) by an independent examiner (M.T.) who was not the operating surgeon (H.R.). This examiner was blinded to the method of tendon attachment and was not involved in the radiographic evaluation.
Thirty-one patients were men and eleven were women. Their mean age at the time of the latissimus dorsi tendon transfer was fifty-eight years (range, fifty to seventy-five years). The right shoulder was affected in twenty-eight patients, and the left was affected in fourteen. The dominant side was affected in thirty patients. The Constant and Murley score14 and the American Shoulder and Elbow Surgeons (ASES) shoulder score15 were used to evaluate the patients clinically. Pain was graded on a scale ranging from 0 to 15 points, with 0 points indicating worst pain and 15 points indicating no pain, according to the method of Constant and Murley14. Abduction strength was measured (in kilograms) in the scapular plane, and external rotation strength was measured in the axial plane with the adducted arm at the side. Strength measurements were performed with an Isobex isometric dynamometer (Cursor, Bern, Switzerland) and compared with the values for the unaffected, asymptomatic shoulder. Additionally, the patients rated their subjective satisfaction with the final outcome on a visual analog scale ranging from 0 (absolutely dissatisfied) to 10 points (maximal satisfaction). All patients were assessed for the hornblower sign12 preoperatively and postoperatively. Subscapularis function was examined with the lift-off test13.
In 2002, we modified our tendon harvesting technique in an effort to yield more stable fixation of the latissimus tendon at the greater tuberosity. This decision was based on our excellent results after using the same harvesting technique for direct transfer of the pectoralis major muscle to achieve dynamic stabilization of the scapula in patients with serratus anterior palsy. The last twenty-two consecutive patients who underwent latissimus dorsi tendon transfer were operated on with use of this modified technique. Two of these patients died of unrelated causes; thus, twenty were included in this study. Two cohorts of patients were evaluated. Group A consisted of twenty-two patients who were operated on, between 1997 and 2001, with a conventional sharp tendon release from the humeral insertion. In ten of the twenty-two patients, the latissimus dorsi tendon transfer was performed as revision surgery after a previous failed rotator cuff repair. Group B included twenty patients who were operated on, between 2002 and 2005, with use of a modified tendon-harvesting and repair technique that involved more stable, bone-to-bone fixation. Eight of the twenty patients had the latissimus dorsi tendon transfer performed as a revision procedure.
To avoid a possible influence of different follow-up periods on the final outcome, patients in Group A were followed for two years after enrollment of the last patient. Thus, the two groups were comparable with regard to follow-up time; they were also comparable in terms of age and sex, duration of symptoms, numbers of previous attempts at rotator cuff repair, and possible negative cofactors such as smoking or the postoperative use of anti-inflammatory drugs.
Radiographic Evaluation
Preoperative and follow-up radiographs in two planes (a true anteroposterior view with the arm in neutral position and an axillary view) were made for all patients. The stage of osteoarthritis was classified with the system of Samilson and Prieto16 on the basis of these radiographs. Superior migration of the humeral head was graded as absent, present, or severe according to the criteria defined by Gerber3. The line joining the inferior scapular border and the calcar of the humerus was defined as the Shenton line of the shoulder. Superior migration of the humeral head was graded as absent if the Shenton line was uninterrupted and the center of the humeral head lay at the center of the glenoid. An interrupted Shenton line indicated the presence of proximal humeral head migration, which was graded as severe if there was obliteration of the subacromial space with associated marked sclerosis of the acromial undersurface. Like Aoki et al.2, we did not measure the acromiohumeral distance to evaluate superior migration of the humeral head. When a supplementary acromioplasty had been performed at the time of the index surgery, this distance was increased iatrogenically; thus, no conclusion could be drawn regarding the depressor activity of the transferred latissimus dorsi.
Magnetic Resonance Imaging
All patients underwent magnetic resonance imaging with T1 and T2-weighted images of the involved shoulder in the sagittal and coronal oblique planes at the final follow-up examination, at which time plain radiographs were also made for this study. As mentioned, the follow-up times were similar for Groups A and B. The main purpose of the magnetic resonance imaging was either to confirm healing of the transferred tendon to the greater tuberosity or to detect avulsion of the tendon. If the tendon was not clearly identified or did not demonstrate direct contact with the greater tuberosity, the radiographic result was rated as a failure. Additionally, atrophy and fatty degeneration of the latissimus dorsi were evaluated on T1-weighted images. Moreover, the integrity of the subscapularis and teres minor tendons was analyzed, as it has been reported that patients with an intact subscapularis and teres minor have a better outcome following latissimus dorsi tendon transfer2,4,17. All magnetic resonance images were reviewed by three independent examiners (M.M., H.R., and R.F. [a musculoskeletal radiologist]) who were blinded to patient data, the operative technique, and the clinical results. Interobserver differences among the three examiners were evaluated with use of an intraclass correlation coefficient18. An intraclass correlation coefficient of 0 to 0.40 was considered to indicate poor interobserver reliability; 0.41 to 0.75, fair or moderate reliability; and >0.75, excellent reliability.
Surgical Technique
The procedure was performed with the patient in the beach-chair position and under general endotracheal anesthesia combined with an interscalene cervical plexus block. A superior transdeltoid approach to the rotator cuff with a skin incision of about 4 to 5 cm was performed to evaluate the rotator cuff. An acromioplasty was performed if acromial spurs were present, and a tenotomy of the long head of the biceps tendon was done if degeneration and fraying of the biceps tendon were seen. When an acromioplasty was carried out, the deltoid was detached from the anterior aspect of the acromion. An attempt at tendon mobilization and rotator cuff repair was made in all patients. When a diagnosis of a massive irreparable posterosuperior defect of the rotator cuff was made, a second skin incision of approximately 8 cm in length was made in the posterior axillary crease with the arm in 90° of abduction. Care was taken to protect the posterior brachial cutaneous nerve. The latissimus dorsi muscle was identified, and its tendon was dissected (Fig. 1). Blunt dissection was used to separate the latissimus dorsi from the teres major muscle, which was difficult in some cases because of a shared, conjoint tendon.
At this point, the standard technique was modified in Group B. Instead of sharp dissection of the tendon from the bone, the tendon insertion was harvested along with a small bone chip with use of a chisel (Fig. 2). The bone fragment encompassed the entire width of the latissimus dorsi tendon and had a diameter of 2 to 3 mm. Two or three number-2 nonabsorbable tagging sutures were placed behind the bone chip in the tendon for retraction (Fig. 3). The neurovascular pedicle was identified and protected. For muscle transfer, a tunnel was created deep to the deltoid and extended to the greater tuberosity by blunt dissection. A long vascular clamp was then passed through the tunnel inferiorly and out through the axillary incision. With use of the stay sutures, the latissimus dorsi tendon was passed through the tunnel and pulled superolaterally to the posterosuperior aspect of the greater tuberosity. At this stage, care must be exercised to avoid rotating the tendon when it is pulled through. Next, an osseous trough was created with a burr in the greater tuberosity as previously described19. Four or five number-2 nonabsorbable sutures were then used to achieve transosseous bone-to-bone fixation of the osseous chip. The same steps were performed in Group A, but the tendon was sharply detached from the humerus and fixed to the greater tuberosity with use of four or five transosseous sutures.
Postoperative Treatment
The patient's arm was immobilized at the side in a sling for four weeks postoperatively. During this time, only passive exercises, including abduction and flexion up to 60°, were permitted under a physiotherapist's guidance. Isometric myoelectrostimulation was performed to stimulate the transposed muscle. After four weeks, active exercises, with avoidance of provocation of pain, were begun in all planes. After twelve weeks, full motion of the shoulder was permitted and the patient was allowed to return to light manual work and sports activities.
Statistical Analysis
Statistical analysis was performed to determine whether the clinical outcome at the time of follow-up represented a significant change in comparison with the preoperative status. The Wilcoxon signed-rank test was used to compare preoperative and postoperative nonparametric data in both groups as well as to compare the final Constant and ASES scores between Group A and Group B. The level of significance was set at p = 0.05.
Source of Funding
No external funds were received for this study.
Clinical Results
In Group A, the mean Constant score improved from 43.4 points (range, 24 to 52 points) preoperatively to 64.8 points (range, 24 to 80 points) postoperatively. The mean ASES score improved from 49.3 points (range, 35 to 56 points) to 69.6 points (range, 42 to 86 points). The mean Constant pain score improved from 5.6 to 11.9 points, whereas subjective satisfaction with the final outcome as recorded on the visual analog scale averaged 7.1 points (range, 2 to 10 points). All of these improvements were significant (p < 0.05).
In Group B, the mean Constant score increased from 40.2 points (range, 24 to 54 points) to 74.2 points (range, 48 to 87 points), and the mean ASES score increased from 47.2 points (range, 35 to 56 points) to 77.1 points (range, 50 to 91 points). The Constant pain score improved from 5.2 to 13.8 points, and the subjective satisfaction as recorded on the visual analog scale averaged 9.2 points (range, 2 to 10 points). These improvements were significant (p < 0.05).
To show the influence of the modified technique of latissimus dorsi tendon transfer, the clinical scores, range of motion, and strength measurements in each group and those in the entire study group of patients who underwent latissimus dorsi tendon transfer are reported separately in Table I. The mean Constant score (64.8 points in Group A and 74.2 points in Group B), ASES shoulder score (69.6 points in Group A and 77.1 points in Group B), range of motion, and strength were significantly better in Group B than in Group A (p < 0.05).
Before the operation, twenty-eight patients showed a positive hornblower sign, which remained positive in five patients after the index procedure.
The mean Constant score for the twenty-four patients treated with a primary operation was significantly better than that for the eighteen treated with a salvage procedure (76.2 compared with 60.3 points; p < 0.05).
Eight (19%) of the forty-two patients had an unsatisfactory result, which was defined as a lack of clinical improvement in the Constant score as compared with the preoperative score. Four patients had a rupture of the transferred tendon, two had secondary spontaneous subscapularis deficiency at two and four years, and the cause for the unsatisfactory outcome was unknown for two patients. The mean Constant score for these eight patients with an unsatisfactory result was 46 points (range, 32 to 54 points) at the time of follow-up. Of the patients with an unsatisfactory outcome, six (27%; four with a rupture of the transferred latissimus tendon, one with secondary spontaneous subscapularis deficiency, and one with an unknown cause) were in Group A and two (10%; one with secondary spontaneous subscapularis deficiency and one with an unknown cause) were in Group B. This difference between the percentages of patients with an unsatisfactory outcome in the two groups was significant (p < 0.05).
Radiographic Results
Superior migration of the humeral head increased between the preoperative and postoperative examinations in twenty-six patients (62%). The grade of the superior migration changed from absent to present in nineteen patients, from absent to severe in three, and from present to severe in four. Four shoulders had a decrease from present to absent, and twelve patients had no change.
Before surgery, ten patients showed signs of glenohumeral osteoarthritis according to the classification system of Samilson and Prieto16; eight of the cases were mild, and two were moderate. Two patients with mild preoperative degenerative changes showed moderate osteoarthritis at the two-year follow-up examination, and one with moderate preoperative changes had severe osteoarthritis at the time of follow-up. Nine patients without osteoarthritic changes before the index procedure showed osteoarthritis at the time of follow-up; six of these cases were mild, and three were moderate.
Comparison of the radiographic results between Groups A and B indicated that the technical modification had no apparent influence on the development of osteoarthritis or superior humeral head migration; however, a statistical comparison of Groups A and B was not performed because of the limited number of patients with osteoarthritic changes.
Magnetic Resonance Imaging Findings
The interobserver reliability of the evaluation of the integrity of the transferred tendon, the course of which could be followed clearly, was graded as excellent, with an intraclass correlation coefficient of 0.79. There were difficulties with the assessment of the surgical technique in the presence of the bone chip at the greater tuberosity. This can be explained by the fact that cortical bone is generally difficult to evaluate on magnetic resonance imaging. Alternatively, there may have been osseous integration of the bone chip. The interobserver reliability of the assessment of bone integration was graded as poor, with an intraclass correlation coefficient of 0.36.
In Group A, four patients showed no healing and had retraction of the latissimus dorsi tendon. Fatty infiltration and muscle atrophy were observed in these four cases. Nevertheless, one patient showed clinical improvement (of 5 points in the Constant score) due to pain relief compared with the preoperative state. Subscapularis deficiency developed in one patient in Group A who had integrity of the latissimus dorsi tendon transfer, and this patient had a poor clinical outcome (a Constant score of 37 points).
In Group B, all patients showed a normal course of the transferred muscle without any sign of avulsion of the tendon from the greater tuberosity. No signs of fatty degeneration or atrophy were seen in the muscle belly of the transferred latissimus dorsi. A rupture of the subscapularis tendon was observed in one patient in this group. The clinical outcome of this patient was rated as poor, with a Constant score of 44 points.
Two patients, one in each group, showed subscapularis deficiency with a complete tendon tear and mild fatty infiltration. According to the system of Goutallier et al.9, fatty infiltration of the teres minor was stage 0 in twelve patients, stage 1 in eleven, stage 2 in eight, stage 3 in nine, and stage 4 in two. The results for the patients with stage-3 or 4 fatty infiltration were significantly inferior to those with stage-0, 1, or 2 infiltration (p = 0.018). We did not perform a statistical comparison of the clinical relevance of fatty infiltration of the teres minor in Groups A and B because of the limited number of patients.
Complications
An axillary neurapraxia with spontaneous remission after six months developed in one patient in Group B. Nonetheless, the final clinical outcome for this patient was excellent, with a Constant score of 81 points.
Latissimus dorsi tendon transfer is a well-established reconstructive procedure for the treatment of irreparable rotator cuff tears involving the supraspinatus and infraspinatus tendons. However, the results must be interpreted within the context of it being a salvage procedure with variable results and a higher failure rate. Several authors have tried to identify factors affecting the outcome. Iannotti et al.6 found female sex, poor preoperative shoulder function, generalized muscle weakness, and the absence of electrical activity of the transferred muscle at the time of follow-up to be negative factors. Warner and Parsons8 reported that latissimus dorsi tendon transfer had better results when it was performed as a primary procedure than when it was done as a salvage procedure after a failed rotator cuff repair. Gerber et al.4 observed a significant positive correlation between the integrity of the subscapularis tendon and the clinical outcome. In a later study, teres minor integrity was identified as a significant negative predictor17 when fatty infiltration exceeded stage 2 according to the system of Goutallier et al.9. Werner et al.20 investigated the influence of innervation and psychomotor skills on the outcome of latissimus dorsi tendon transfer and concluded that the function of the operatively treated shoulder could be predicted by the psychomotor function of the uninjured, contralateral shoulder. All of these studies and various observations emphasize the multifactorial nature of the outcome of latissimus dorsi tendon transfer.
We hypothesized that rupture of the tendon from its new humeral insertion on the greater tuberosity could be responsible for a substantial number of the unsatisfactory results reported after latissimus dorsi tendon transfer in the literature, and this hypothesis was confirmed by the results of the current study. Magnetic resonance imaging showed detachment of the tendon in four of the patients who had an unsatisfactory clinical result. All four of these patients were in Group A, and they accounted for two-thirds of the patients with an unsatisfactory result in this group. The main purpose of the technical modification of the latissimus dorsi tendon transfer was to improve the structural integrity of the transferred tendon, which often was found to consist of thin and insubstantial tissue. We postulated that bone-to-bone healing could be achieved by including a small amount of cortical bone with the latissimus dorsi tendon and that it would improve fixation stability, thereby reducing the subsequent rate of failure.
Muscular balancing is crucial for the success of a latissimus dorsi tendon transfer. Our results support the findings of Gerber3 and Aoki et al.2 that an insufficient subscapularis tendon negatively influences the final outcome. In our series, two patients with an intact, healed latissimus dorsi tendon but a poor clinical result at the time of follow-up showed deficiency of the subscapularis tendon on both the clinical and the magnetic resonance imaging examination. Secondary spontaneous rupture of the subscapularis tendon occurred in these cases. Werner et al.20 attributed the inferior results of latissimus dorsi tendon transfer in the presence of subscapularis dysfunction to the loss of centering of the humeral head on abduction and elevation.
Fatty infiltration of the teres minor has been reported to play a key role in the outcome of latissimus dorsi tendon transfer17. Our observations confirm the crucial role of teres minor integrity, as the patients with extensive fatty infiltration of the teres minor muscle had inferior results. In this series, teres minor deficiency was not considered to be a contraindication to latissimus dorsi tendon transfer because its negative prognostic value was unknown. Therefore, several patients with either clinically evident teres minor deficiency (a positive hornblower sign) or pathological changes in the teres minor on magnetic resonance imaging were treated with the procedure. However, our results suggest that preoperative teres minor deficiency with advanced fatty infiltration should definitely be considered a negative prognostic factor and as a contraindication to latissimus dorsi tendon transfer.
The results of the latissimus dorsi tendon transfers that were done as revision procedures were inferior to the results of the primary procedures. This finding confirms the observations by Gerber et al.4 and Warner and Parsons8. Possible reasons for this finding were the inferior preoperative shoulder function and the advanced age of the patients with a previous failed rotator cuff repair compared with the patients undergoing the tendon transfer as a primary procedure. For instance, six of the eighteen patients who underwent latissimus dorsi tendon transfer as a revision procedure had a longstanding severe pseudoparesis of elevation, which is known to be a negative factor21.
Aoki et al.2 observed progression of glenohumeral osteoarthritis in 41% of their cases, and Gerber et al.4 observed it in 30%. These observations were confirmed in the present study by a progression rate of 29%, independent of the surgical technique that had been used. Nevertheless, patients obtain a clear short-term benefit from this procedure, and mild preoperative osteoarthritis does not appear to be a contraindication to latissimus dorsi tendon transfer even if the osteoarthritic changes increase slightly over time.
One of the major limitations of this study results from its retrospective design with comparison of two nonconcurrent, highly variable patient populations in which there were multiple factors for failure. The new surgical technique was applied after surgical experience with latissimus dorsi tendon transfer had been gained in the first group of twenty-two patients. This may have influenced, in a positive manner, the outcomes in the second cohort. The lack of electromyographic information can be seen as another limitation of the study because an analysis of the functional activity of the transferred muscle was not possible. We performed electromyography only for patients with clinical failure, to identify the reasons for the failure. Furthermore, as a result of the small number of patients, there was insufficient power to allow a statistical analysis of the importance of small differences between the two study groups. In addition, the functional outcome of the latissimus dorsi tendon transfer was compared with the status of the unaffected shoulder. It must be remembered that a certain percentage of these patients probably had pathological changes in the rotator cuff of the contralateral shoulder, which have may influenced the isometric measurements. However, all patients in this study had a lack of symptoms in the unaffected shoulder at the time of follow-up.
Overall, the outcome of latissimus dorsi tendon transfer is influenced negatively by a series of associated factors, including subscapularis and teres minor deficiency, patient age, preoperative shoulder function, and a previous failed rotator cuff repair. The modification of tendon harvesting to include a small piece of bone, as described in this study, can reduce the rate of one factor, rerupture of the transferred tendon. 