Tears of the rotator cuff tendons are very common. Uhthoff et al.1 found a 20% prevalence in a series of cadaver dissections in which the mean age of the donors was 59.4 years. Lehman et al.2 found a prevalence of 17% in a large series of cadaver dissections, with a prevalence of 30% in donors older than sixty years of age. Surgical repair of the rotator cuff is the preferred treatment following unsuccessful conservative therapy. Studies of open repair of the rotator cuff have shown that function and strength can be successfully restored and that pain is effectively relieved3-6. Complications related to open rotator cuff repair include infection, failure of the deltoid muscle reattachment, and retearing of the rotator cuff7. Failure of the rotator cuff repair is the most frequently observed complication, with estimates of the rate ranging between 20% and 94%8,9. Failure of the repair of the rotator cuff itself may occur secondary to poor tendon or bone quality, failure of suture or knots, inadequate fixation of tendon to bone, lack of tendon-to-bone healing, or inappropriate postoperative care10-12.
The most common approach to rotator cuff repair using an arthroscopic technique involves the use of suture anchors in either a single-row or a double-row configuration. The single-row repair involves placing anchors either in the lateral aspect of the tendon footprint or lateral to the footprint itself13,14. A double-row repair incorporates the same anchor configuration as the single-row repair, with the addition of a second row of anchors placed in the medial aspect of the tendon footprint15,16.
Studies have demonstrated good clinical outcomes following arthroscopic single-row repair13,17. However, techniques have evolved to include a double row in an effort to improve healing rates15,16. A number of basic science studies have compared the two techniques and have demonstrated superiority of fixation strength in double-row compared with single-row repair18-22. The present study used a disease-specific quality-of-life outcome measure to evaluate the postoperative improvement and to compare the outcomes between groups.
The primary research objective was to determine whether patients who undergo a repair of the rotator cuff with an arthroscopic technique involving double-row fixation have improved disease-specific quality of life, as measured by the Western Ontario Rotator Cuff Index (WORC) at two years postoperatively, compared with patients who undergo a repair involving single-row lateral fixation. Secondary research objectives included determination of differences in outcome between the two groups as measured by the Constant score, the American Shoulder and Elbow Surgeons (ASES) score, and strength. The healing rate was determined at one year postoperatively with use of ultrasonography or magnetic resonance imaging (MRI).
This was a multicenter, double-blind, balanced (1:1) randomization study conducted at two orthopaedic surgery facilities: The Ottawa Hospital in Ottawa, Ontario, and the Pan Am Clinic in Winnipeg, Manitoba, Canada. Enrollment occurred from June 2007 to June 2009. The study received approval by the Research Ethics Boards at both institutions. Patients were recruited in the outpatient specialty clinic setting by the study coordinator or research assistant. The target population was men and women of any age with a diagnosis of a full-thickness tear of the rotator cuff according to clinical criteria (including MRI in all patients). All patients had continued to have persistent pain and functional disability after at least six months of conservative treatment (including activity modification, administration of analgesic or anti-inflammatory medication, and/or physiotherapy) ordered by the referring physician. Rotator cuff tears of any size were considered for inclusion. Patients who agreed to be contacted were approached by one of two research assistants who conducted the study enrollment process.
Exclusion criteria were (1) an acute rotator cuff tear (a clear history of traumatic etiology with symptoms present for less than six months), (2) characteristics of the tear that rendered the rotator cuff irreparable (e.g., fatty infiltration in the muscles that was 50% [grade III in the classification described by Goutallier et al.10,23] or greater, as measured by MRI), (3) superior subluxation of the humeral head (an acromiohumeral distance of <7 mm, as measured on a true anteroposterior radiograph with the arm in neutral rotation), (4) a substantial shoulder comorbidity (e.g., a Bankart lesion or osteoarthritis [defined as joint space narrowing or the presence of osteophytes visible on radiographs and confirmed by arthroscopy]), (5) previous surgery on the affected shoulder, (6) an active Workers’ Compensation claim, (7) active joint or systemic infection, (8) rotator cuff tear arthropathy, (9) a substantial medical comorbidity that could alter the effectiveness of the surgical intervention, (10) a major medical illness (resulting in a life expectancy of less than one year or an unacceptably high operative risk), (11) inability to speak or read English or French, (12) a psychiatric illness that precluded informed consent, and (13) unwillingness to be followed for two years.
The primary outcome measure was the WORC, a disease-specific quality-of-life measure for rotator cuff disease24. Secondary outcome measures included the ASES score25, the Constant score26, and shoulder strength in forward elevation. Details of the clinical outcome instruments and their administration are summarized in the Appendix. Additional secondary outcomes monitored during the postoperative course included complications and the incidence of revision surgery.
Nonarthrographic MRI scans were obtained preoperatively in all patients and were used to determine the initial tear size in both the coronal oblique and the sagittal oblique plane. The method used to determine the tear size was adapted from Bryant et al.27. Measurement of the long-axis dimension was made from the lateral aspect of the torn tendon to the lateral extent of the footprint insertion on the oblique coronal plane images. The short-axis measurement was made from the anterior to the posterior tendinous extent of the tear on the oblique sagittal plane images. Ultrasonography or MRI was used to determine the healing rate at twelve months postoperatively. If the tendons were in continuity with no evidence of full-thickness tearing, the repair was considered healed (intact). A fellowship-trained musculoskeletal radiologist reviewed all imaging studies.
After consent had been received and prior to surgery, the research assistants obtained the functional and quality-of-life outcome data in the clinic setting. Patients completed all outcome measures preoperatively and at three, six, twelve, and twenty-four months postoperatively.
Final eligibility of participants for the study was established on the basis of intraoperative visual inspection of the rotator cuff tear and determination of reparability. A full-thickness tear was classified as repairable if the tendon could be restored to its anatomical insertion when traction was applied without undue tension. All surgical procedures were performed by one of two fellowship-trained shoulder surgeons (P.L.C.L. and P.M.).
Once eligibility was confirmed, patients were randomized to one of two methods of arthroscopic rotator cuff fixation, single-row or double-row repair. Allocation was carried out with use of computer-generated block randomization; the surgeon was blinded to the block size. The treatment allocations were printed on cards and inserted into sealed, opaque envelopes. The circulating nurse opened an envelope once patient eligibility was confirmed.
Repair of the rotator cuff tear was carried out with use of the assigned technique, either single-row or double-row fixation using an arthroscopic method (Fig. 1). Both procedures were performed under general anesthesia with the patient in either the beach-chair or the lateral position. Prophylactic antibiotics (cefazolin, or clindamycin in the case of penicillin allergy) were administered. An examination under anesthesia was performed and documented. All patients underwent a standard diagnostic arthroscopy. All shoulder pathology encountered was documented on a standardized form. The surgical repair was then performed using a standardized technique. Super Revo anchors (ConMed Linvatec, Largo, Florida) were used at the Ottawa site, and Duet anchors (ConMed Linvatec) were used at the Winnipeg site. Both anchors are of the threaded, “screw-in” variety and are similar in size; they differ only in composition, with Super Revo anchors being all-metal and Duet anchors being composed of PLLA (poly-L-lactic acid). The number of anchors used was determined at the discretion of the surgeon on the basis of the size and complexity of the tear.
Single-row fixation involved fixation with suture anchors double-loaded with number-2 high-tensile-strength sutures placed in a single row parallel to the sagittal axis of the rotator cuff footprint. Anchors were placed along the lateral edge of the footprint. Mattress or inverted-mattress sutures were used, with sliding-locking knots and alternating half-hitches.
Double-row fixation involved fixation with suture anchors placed along two rows. The lateral-row fixation was the same as that described for single-row fixation. The medial-row anchor(s) were double-loaded with number-2 high-tensile-strength sutures and were parallel to the sagittal axis of the rotator cuff footprint. Anchors were placed along the bone-cartilage junction at the medial aspect of the footprint. Mattress sutures were used, with sliding-locking knots and alternating half-hitches. The anchor sutures in the medial row were not tied or linked to the lateral anchor sutures or lateral anchors.
Patients were discharged on the same day as the surgery. Pendulum exercises were initiated on the first postoperative day. Active-assisted shoulder motion exercises were initiated six weeks postoperatively. Active motion began between eight and twelve weeks postoperatively, with strengthening exercises and reintegration into normal activities starting at twelve weeks postoperatively. Patients returned to the clinic at two and six to eight weeks postoperatively for wound checks only, and they returned at three, six, twelve, and twenty-four months postoperatively for study follow-up. Research assistants not involved in the surgical procedure conducted the administration of all of the outcome measures in the clinic setting.
Because of the nature of a surgical trial, it was not possible to blind the surgeon to the intervention. However, the trained research assistant who performed the follow-up assessments did not have prior access to the patient’s chart and was therefore blinded to the treatment assignment. The patient was also blinded to the treatment assignment.
MRI or ultrasonography was performed at a minimum of twelve months postoperatively to assess healing. An independent investigator who was a fellowship-trained musculoskeletal radiologist evaluated all of the images.
This trial was registered at www.clinicaltrials.gov (trial number NCT00508183).
Statistical Methods
A sample size calculation was performed on the basis of the mean WORC score (and standard deviation) of 71.31 ± 30.43 obtained in a previous study of single-row fixation. We assumed that the outcome for double-row fixation would be superior and determined the sample size required to detect a difference of 30%, or 21.4 points, between the mean WORC scores of the two groups. With an anticipated dropout rate of 20%, forty-two patients per group would therefore be needed.
Baseline demographics were compared between the two groups with use of a t test for continuous variables and a chi-square test for categorical variables. A Wilcoxon rank sum test was used for comparing the number of anchors. Values in each group and in the overall cohort are presented as the mean, standard deviation, and range
Longitudinal analysis using mixed-effect modeling, with patients as random effects, was used to assess whether the WORC, ASES, and Constant scores and strength improved significantly in each group over time. A two-sided t test was used to compare each outcome between the two treatment groups at baseline, three months, six months, one year, and two years. The healing rates were compared between the two treatment groups with use of a chi-square test. Outcome scores were also compared between patients with healed and retorn rotator cuff repairs with use of a two-sided t test. Multivariate logistic regression was performed to determine the association of selected variables with healing status (intact or retorn rotator cuff); these variables included randomization group (single or double row), tear size, age, number of anchors, and baseline outcome scores.
A p value (alpha) of 0.05 was considered significant. All analyses were performed with use of SAS software (version 9.1; SAS Institute, Cary, North Carolina).
Source of Funding
There was no external funding source for this study.
The flow of patients through the study is summarized in Figure 2. A total of 118 patients were enrolled. Nine of these patients were never randomized because they either postponed or canceled the surgery, and nineteen patients were excluded prior to randomization for other reasons. The remaining ninety patients were randomized to receive either single-row or double-row fixation. Seventeen of these patients were lost to follow-up by the twenty-four-month outcome assessment, yielding 81% retention of patients in the trial by the time of final follow-up.
Baseline demographic data are summarized in Table I. Forty-two patients were allocated to the treatment (double-row) arm and forty-eight were allocated to the control (single-row) arm. Patient enrollment and randomization was stopped after the ninetieth patient was randomized. As a result, blocks were left open at both sites, and this accounts for the slightly uneven group assignments. The mean age (and standard deviation) of the study participants was 56.8 ± 8.1 years (range, thirty-eight to eighty-two years), and 71% were men. Baseline characteristics, including age (p = 0.29), sex (p = 0.68), affected side (p = 0.39), and tear size (p = 0.28 for the coronal plane and p = 0.99 for the sagittal plane) did not differ significantly between the groups. The number of anchors was significantly greater in the double-row group (p < 0.0001).
No significant differences between the single-row and double-row groups were found in the WORC, ASES, or Constant scores or strength at any time point (Table II). All of these outcomes improved significantly in both groups from baseline to the time of final follow-up (p < 0.001). However, the single-row and double-row groups differed significantly with regard to the relative improvement over time in the WORC score (p = 0.062 for the improvement between six and twelve months in the single-row group compared with p = 0.003 in the double-row group), the ASES score (p < 0.0001 between baseline and three months in the single-row group compared with p = 0.341 in the double-row group, and p = 0.381 between six and twelve months in the single-row group compared with p = 0.0032 in the double-row group), the Constant score (p = 0.94 between baseline and three months in the single-row group compared with p = 0.078 in the double-row group, and p = 0.148 between twelve and twenty-four months in the single-row group compared with p = 0.01 in the double-row group), and strength (p = 0.913 between six and twelve months in the single-row group compared with p = 0.015 in the double-row group).
Postoperative imaging was performed in seventy-six (84%) of the ninety patients in the study cohort (thirty-nine in the single-row group and thirty-seven in the double-row group). The imaging consisted of ultrasonography in sixty-five of these patients and MRI in the other eleven. Demographic variables did not differ significantly between the patients who returned for postoperative imaging and those who did not (p = 0.215 for age, p = 0.98 for sex, p = 0.136 for side, p = 0.81 for coronal tear size, p = 0.34 for sagittal tear size, and p = 0.93 for the number of anchors).
The healing rate was 67% in the single-row group compared with 78% in the double-row group (p = 0.254). The estimated relative risk of a retear in the double-row group was 0.65 (95% confidence interval 0.30 to 1.38) compared with the single-row group, which was not significant.
The final outcome scores were compared according to the healing status (Table III). Fifty-five patients in the cohort had an intact (fully healed) tendon, and twenty-one had a retear. The mean outcome was significantly poorer in the patients with a retear as assessed with the Constant score (p = 0.044) and strength (p = 0.001). The WORC score (p = 0.22) and the ASES score (p = 0.18) did not differ significantly between the groups. Patients with a retear had a larger initial tear (as measured by MRI) in the coronal plane (p = 0.0018) and the sagittal plane (p = 0.016) compared with patients whose tendon had healed.
The results of the multivariate logistic regression are summarized in Table IV. Two variables, smaller coronal tear size (p = 0.011) and randomization to double-row fixation (p = 0.037), were significantly associated with having a healed rotator cuff on imaging when adjusted for age, sagittal tear size, number of anchors, and baseline outcome scores. No other variables had a significant association with the healing status.
One participant in the single-row group and three in the double-row group required further surgery; this difference was not significant (p = 0.335). All four of these patients underwent a repeat arthroscopic rotator cuff repair by the twenty-four-month time point. No complications related to surgery occurred during the study.
The disease-specific quality of life at two years postoperatively, as measured by the WORC score, did not differ significantly between the patients who underwent arthroscopic rotator cuff repair using single-row lateral fixation and the patients in which double-row fixation was used. Furthermore, there were no differences between the groups in the secondary measures of ASES and Constant scores and strength. Therefore, our hypothesis that double-row fixation yields superior quality-of-life outcomes compared with single-row lateral fixation was not supported. The healing rate did not differ significantly between the single-row and double-row groups. However, multivariate regression analysis that controlled for other variables including age, number of anchors, and baseline outcome scores demonstrated that both a smaller initial tear size and double-row repair were associated with a greater healing rate.
Limitations of the study include a loss to follow-up of 19%. However, the sample size calculation accounted for a 20% loss to follow-up. In the surgical technique, either mattress or inverted mattress sutures were used for the lateral-row fixation. However, it seems unlikely that this variable would have affected the results, given that both suture configurations appear to trap approximately the same amount of tendon tissue. Sixteen percent of the patients did not agree to return for postoperative imaging, and there is a potential for bias as a result of this loss to follow-up. The imaging differed between the two centers; ultrasonography was used at the Ottawa site and MRI was used at the Winnipeg site. However, as both modalities have demonstrated high sensitivity for the detection of rotator cuff tears28,29, including in the postoperative setting30, we do not believe that this substantially weakens the quality of the data. Another limitation was related to the two-year duration of follow-up. It is possible that longer follow-up would have been of value. However, as soft-tissue healing can be considered to be complete by twelve months31,32, it seems likely that any further changes in healing status would be related to chronic intrinsic tendon pathology, which should be similar between groups. Furthermore, in a long-term study, Jost et al. demonstrated that functional outcomes in patients with a nonhealed rotator cuff repair were maintained at 7.6 years following repair33.
The rate of nonhealing after rotator cuff repair has been reported to be surprisingly high, between 20% and 94%8,9. This rate of failure of the tendon to heal following surgical repair has spurred further research into surgical repair technique. The surface area provided by different types of repairs was studied by Apreleva et al.34. The authors quantitatively studied the three-dimensional area of the original supraspinatus insertion and compared this area with that achieved by four different methods of reconstruction in ten human cadavers. The methods included transosseous simple suturing, transosseous mattress suturing, suture-anchor simple suturing, and suture-anchor mattress suturing. The authors concluded that transosseous simple suturing provided the greatest area, 85% of the original surface area, whereas the other three methods reproduced only 67% of the original area.
If the area of the anatomic rotator cuff insertion is not reproduced, the results are likely to vary depending on the bone quality at the insertion site. Uhthoff et al.35 demonstrated that the portion of tendon in contact with the osseous trough influences the quality of healing in a rabbit model. Fibrocartilage formed at sites where the transected tendon was in contact with bone but not where the articular or bursal tendon surface contacted bone.
As single-row repair only partially restores the native tendon-to-bone insertion (footprint), a double-row technique to maximize contact area was described15,16. Kim et al. compared single-row with double-row fixation in nine pairs of cadaver shoulders18. The authors reported that double-row repair demonstrated superior performance compared with single-row repair. Gap formation, strain, repair stiffness, and ultimate load to failure were superior in the double-row repair. However, results have not been consistent among cadaveric studies. Mazzocca et al. compared four arthroscopic repair techniques in twenty human cadaveric shoulders19. The authors compared single-row fixation with three types of double-row fixation: diamond, mattress double anchor, and modified mattress double anchor. Footprint area, displacement under cyclic loading, and load to failure were examined. A significantly larger cuff footprint area was found in all double-row configurations, but no differences between single-row and double-row fixation were found in rotator cuff displacement or load to failure.
A study by Franceschi et al. provided Level-I clinical evidence36. The authors studied fifty-two patients randomized to either single-row or double-row fixation. No difference in the University of California Los Angeles (UCLA) score or shoulder motion was observed, and no difference in rotator cuff repair integrity was observed with use of MRI. Grasso et al. compared the two repair methods in eighty patients in a randomized clinical trial37. The authors did not observe any differences between groups in the Disabilities of the Arm, Shoulder and Hand (DASH) score, Constant score, or strength. Burks et al. randomized forty patients to single-row or double-row rotator cuff repair38. There were two retears in each group at one year. The authors did not observe any differences between the groups in clinical outcomes or in healing rates assessed with MRI.
The results of the present study are in agreement with the other randomized clinical trials comparing single-row and double-row fixation. The quality-of-life outcome scores, functional outcome scores, and strength did not differ significantly between the single-row and double-row repair groups at any time point. However, our observation of an association between double-row fixation and a greater healing rate differs from the results of the other Level-I studies. It is interesting that the proportion of patients requiring a repeat rotator cuff repair did not differ significantly between the groups despite this finding. Analysis of outcome scores according to healing status demonstrated that patients with an intact repair had higher Constant scores and strength measurements. This finding and our observation that tendon retears occurred more commonly in patients with a larger initial tear are consistent with the results of the study by Boileau et al.13.
The degree of improvement across the various time points was not completely uniform between the single-row and double-row groups. A pattern favoring the double-row group with regard to improvement between six and twelve months in the WORC and ASES scores and strength was observed, but the importance of this observation is unclear.
Potential negative aspects of double-row fixation include greater implant cost and longer operative time, greater technical difficulty, occupancy of more space in the proximal aspect of the humerus by anchors, and the theoretical potential for increased pain because of the increased number of anchors.
In summary, the results of this trial did not demonstrate significant differences in functional or quality-of-life outcomes between single-row and double-row repair. However, smaller tears and double-row fixation were associated with greater healing rates of the rotator cuff.